Introductory Statistics, Third Edition, presents statistical concepts and techniques in a manner that will teach student

*383*
*112*
*8MB*

*English*
*Pages 818
[842]*
*Year 2010*

*Table of contents : Cover......Page 1Introductory Statistics......Page 5Copyright......Page 6Table of Contents......Page 9Dedication......Page 8About the Author......Page 7Preface......Page 19Acknowledgments......Page 23 1.1 Introduction......Page 25 1.2 The Nature of Statistics......Page 27 1.3 Populations and Samples......Page 29 1.4 A Brief History of Statistics......Page 31 Key Terms......Page 34 Review Problems......Page 35Chapter 2. Describing Data Sets......Page 41 2.2 Frequency Tables and Graphs......Page 42 2.3 Grouped Data and Histograms......Page 56 2.4 Stem-and-Leaf Plots......Page 68 2.5 Sets of Paired Data......Page 75 2.6 Some Historical Comments......Page 82 Key Terms......Page 83 Summary......Page 84 Review Problems......Page 87Chapter 3. Using Statistics to Summarize Data Sets......Page 95 3.1 Introduction......Page 96 3.2 Sample Mean......Page 97 3.3 Sample Median......Page 107 3.4 Sample Mode......Page 121 3.5 Sample Variance and Sample Standard deviation......Page 123 3.6 Normal Data Sets and the Empirical Rule......Page 133 3.7 Sample Correlation Coefficient......Page 144 Key Terms......Page 158 Summary......Page 160 Review Problems......Page 162Chapter 4. Probability......Page 169 4.2 Sample Space and Events of an Experiment......Page 170 4.3 Properties of Probability......Page 177 4.4 Experiments Having Equally Likely Outcomes......Page 185 4.5 Conditional Probability and Independence......Page 191 *4.6 Bayes' Theorem......Page 209 *4.7 Counting Principles......Page 213 Key Terms......Page 222 Summary......Page 224 Review Problems......Page 225Chapter 5. Discrete Random Variables......Page 233 5.1 Introduction......Page 234 5.2 Random Variables......Page 235 5.3 Expected Value......Page 242 5.4 Variance of Random Variables......Page 255 5.5 Binomial Random Variables......Page 262 *5.6 Hypergeometric Random Variables......Page 272 *5.7 Poisson Random Variables......Page 274 Summary......Page 278 Review Problems......Page 280Chapter 6. Normal Random Variables......Page 285 6.2 Continuous Random Variables......Page 286 6.3 Normal Random Variables......Page 290 6.4 Probabilities Associated with a Standard Normal Random Variable......Page 295 6.5 Finding Normal Probabilities: Conversion to the Standard Normal......Page 301 6.6 Additive Property of Normal Random Variables......Page 303 6.7 Percentiles of Normal Random Variables......Page 308 Summary......Page 314 Review Problems......Page 317Chapter 7. Distributions of Sampling Statistics......Page 321 7.2 Introduction......Page 322 7.3 Sample Mean......Page 323 7.4 Central Limit Theorem......Page 328 7.5 Sampling Proportions from a Finite Population......Page 337 7.6 Distribution of the Sample Variance of a Normal Population......Page 347 Key Terms......Page 349 Summary......Page 350 Review Problems......Page 351Chapter 8. Estimation......Page 355 8.1 Introduction......Page 356 8.2 Point Estimator of a Population Mean......Page 357 8.3 Point Estimator of a Population Proportion......Page 360 8.4 Estimating a Population Variance......Page 366 8.5 Interval Estimators of the Mean of a Normal Population with Known Population Variance......Page 371 8.6 Interval Estimators of the Mean of a Normal Population with Unknown Population Variance......Page 383 8.7 Interval Estimators of a Population Proportion......Page 395 Key Terms......Page 404 Summary......Page 405 Review Problems......Page 407Chapter 9. Testing Statistical Hypotheses......Page 411 9.2 Hypothesis Tests and Significance Levels......Page 412 9.3 Tests Concerning the Mean of a Normal Population: Case of Known Variance......Page 418 9.4 The t Test for the Mean of a Normal Population: Case of Unknown Variance......Page 433 9.5 Hypothesis tests Concerning Population Proportions......Page 445 Summary......Page 457 Review Problems and Proposed Case Studies......Page 461Chapter 10. Hypothesis Tests Concerning Two Populations......Page 467 10.1 Introduction......Page 468 10.2 Testing Equality of Means of Two Normal Populations: Case of Known Variance......Page 470 10.3 Testing Equality of Means: Unknown Variances and Large Sample Sizes......Page 477 10.4 Testing Equality of Means: Small-Sample Tests When the Unknown Population Variances Are Equal......Page 487 10.5 Paired-Sample t Test......Page 495 10.6 Testing Equality of Population Proportions......Page 505 Summary......Page 517 Review Problems......Page 522Chapter 11. Analysis of Variance......Page 527 11.1 Introduction......Page 528 11.2 One-Factor Analysis of Variance......Page 529 11.3 Two-Factor Analysis of Variance: Introduction and Parameter Estimation......Page 538 11.4 Two-Factor Analysis of Variance: Testing Hypotheses......Page 544 11.5 Final Comments......Page 553 Summary......Page 554 Review Problems......Page 557Chapter 12. Linear Regression......Page 561 12.1 Introduction......Page 563 12.2 Simple Linear Regression Model......Page 564 12.3 Estimating the Regression Parameters......Page 568 12.4 Error Random Variable......Page 577 12.5 Testing the Hypothesis that ß = 0......Page 581 12.6 Regression to the Mean......Page 588 12.7 Prediction Intervals for Future Responses......Page 597 12.8 Coefficient Of Determination......Page 602 12.9 Sample Correlation Coefficient......Page 606 12.10 Analysis of Residuals: Assessing the Model......Page 608 12.11 Multiple Linear Regression Model......Page 610 Summary......Page 619 Review Problems......Page 623Chapter 13. Chi-Squared Goodness-of-Fit Tests......Page 629 13.1 Introduction......Page 630 13.2 Chi-Squared Goodness-of-Fit Tests......Page 633 13.3 Testing for in Dependence in Populations Classified According to Two Characteristics......Page 644 13.4 Testing for Independence in Contingency Tables with Fixed Marginal Totals......Page 655 Key Terms......Page 661 Summary......Page 662 Review Problems......Page 664Chapter 14. Nonparametric Hypotheses Tests......Page 671 14.2 Sign Test......Page 672 14.3 Signed-Rank Test......Page 681 14.4 Rank-Sum Test for Comparing Two Populations......Page 691 14.5 Runs Test For Randomness......Page 700 14.6 Testing the Equality of Multiple Probability Distributions......Page 707 14.7 Permutation Tests......Page 713 Summary......Page 717 Review Problems......Page 720Chapter 15. Quality Control......Page 723 15.2 The X Control Chart for Detecting a Shift in the Mean......Page 724 15.3 Control Charts for Fraction Defective......Page 739 15.4 Exponentially Weighted Moving-Average Control Charts......Page 741 15.5 Cumulative-Sum Control Charts......Page 746 Summary......Page 749 Review Problems......Page 750Appendices......Page 751 Appendix A. Data Set......Page 753 B.2 Absolute Value......Page 757 B.3 Set Notation......Page 758Appendix C. How to Choose a Random Sample......Page 759 Table D.1 Standard Normal Probabilities......Page 763 Table D.2 Percentiles tn,a of t Distributions......Page 764 Table D.3 Percentiles .2n,a of the Chi-Squared Distributions......Page 765 Table D.4 Percentiles of F Distributions......Page 767 Table D.5 Binomial Distribution Function......Page 773Appendix E. Programs......Page 779Answers to Odd-Numbered Problems......Page 781Index......Page 831*

Introductory Statistics

This page intentionally left blank

Introductory Statistics Third Edition

Sheldon M. Ross University of Southern California

AMSTERDAM • BOSTON • HEIDELBERG • LONDON NEW YORK • OXFORD • PARIS • SAN DIEGO SAN FRANCISCO • SINGAPORE • SYDNEY • TOKYO Academic Press is an imprint of Elsevier

Academic Press is an imprint of Elsevier 30 Corporate Drive, Suite 400, Burlington, MA 01803, USA 525 B Street, Suite 1900, San Diego, California 92101-4495, USA Elsevier, The Boulevard, Langford Lane, Kidlington, Oxford, OX5 1GB, UK Copyright © 2010, Elsevier Inc. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this ﬁeld are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. Library of Congress Cataloging-in-Publication Data Ross, Sheldon M. Introductory statistics / Sheldon M. Ross. – 3rd ed. p. cm. Includes bibliographical references and index. ISBN 978-0-12-374388-6 (hardcover : alk. paper) 1. Mathematical statistics. I. Title. QA276.R684 2010 519.5–dc22 2009050832 British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. ISBN: 978-0-12-374388-6

For information on all Academic Press publications visit our Web site at www.elsevierdirect.com

Typeset by: diacriTech, India Printed in Canada 09 10 11 9 8 7 6 5 4 3 2 1

About the Author

Sheldon M. Ross Sheldon M. Ross received his Ph.D. in Statistics at Stanford University in 1968 and then joined the Department of Industrial Engineering and Operations Research at the University of California at Berkeley. He remained at Berkeley until Fall 2004, when he became the Daniel J. Epstein Professor of Industrial and Systems Engineering in the Daniel J. Epstein Department of Industrial and Systems Engineering at the University of Southern California. He has published many technical articles and textbooks in the areas of statistics and applied probability. Among his texts are A First Course in Probability (eighth edition), Introduction to Probability Models (tenth edition), Simulation (fourth edition), and Introduction to Probability and Statistics for Engineers and Scientists (fourth edition). Professor Ross is the founding and continuing editor of the journal Probability in the Engineering and Informational Sciences. He is a fellow of the Institute of Mathematical Statistics and a recipient of the Humboldt U.S. Senior Scientist Award.

v

For Rebecca and Elise

Contents

ABOUT THE AUTHOR....................................................................... v PREFACE.......................................................................................... xvii ACKNOWLEDGMENTS ..................................................................... xxi CHAPTER 1

Introduction to Statistics ............................................

1

1.1 1.2

Introduction ........................................................... The Nature of Statistics ...........................................

1 3

1.2.1

Data Collection ............................................

3

1.2.2 Inferential Statistics and Probability Models ... Populations and Samples .........................................

4 5

*1.3.1 Stratiﬁed Random Sampling .......................... 1.4 A Brief History of Statistics ...................................... Key Terms...................................................................... The Changing Deﬁnition of Statistics................................. Review Problems ............................................................

6 7 10 11 11

Describing Data Sets ....................................................

17

2.1 2.2

18 18

1.3

CHAPTER 2

Introduction ........................................................... Frequency Tables and Graphs .................................. 2.2.1

Line Graphs, Bar Graphs, and Frequency Polygons.....................................................

19

2.2.2

Relative Frequency Graphs ...........................

21

2.2.3

Pie Charts...................................................

24

2.3

Problems ............................................................... Grouped Data and Histograms .................................

25 32

2.4

Problems ............................................................... Stem-and-Leaf Plots ................................................

39 44

Problems ...............................................................

47

* The asterisk signiﬁes optional material not used in the sequel.

vii

viii

Contents

2.5

CHAPTER 3

Sets of Paired Data..................................................

51

Problems ............................................................... 2.6 Some Historical Comments ...................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

54 58 59 60 63

Using Statistics to Summarize Data Sets ......................

71

3.1 3.2

72 73

Introduction ........................................................... Sample Mean ......................................................... 3.2.1

Deviations ..................................................

78

Problems ............................................................... Sample Median.......................................................

79 83

Problems ...............................................................

86

3.4

3.3.1 Sample Percentiles....................................... Sample Mode .........................................................

90 97

3.5

Problems ............................................................... Sample Variance and Sample Standard Deviation .......

98 99

3.6

Problems ............................................................... 105 Normal Data Sets and the Empirical Rule ................... 109

3.7

Problems ............................................................... 114 Sample Correlation Coefﬁcient ................................. 120

3.3

Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

CHAPTER 4

128 134 136 138

Probability ................................................................... 145 4.1 4.2

Introduction ........................................................... 146 Sample Space and Events of an Experiment ............... 146

4.3

Problems ............................................................... 150 Properties of Probability .......................................... 153

4.4

Problems ............................................................... 156 Experiments Having Equally Likely Outcomes ........... 161

4.5

Problems ............................................................... 164 Conditional Probability and Independence................. 167

*4.6

Problems ............................................................... 177 Bayes’ Theorem...................................................... 185 Problems ............................................................... 187

Contents

*4.7

Counting Principles ................................................ 189

Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

CHAPTER 5

Discrete Random Variables .......................................... 209 5.1 5.2

Introduction ........................................................... 210 Random Variables................................................... 211

5.3

Problems ............................................................... 215 Expected Value ...................................................... 218 5.3.1

5.4

5.5

Properties of Expected Values ....................... 221

Problems ............................................................... 225 Variance of Random Variables.................................. 231 5.4.1

Properties of Variances ................................ 233

Problems ............................................................... 236 Binomial Random Variables ..................................... 238 5.5.1

Expected Value and Variance of a Binomial Random Variable ......................................... 243

*5.6

Problems ............................................................... 244 Hypergeometric Random Variables ........................... 248

*5.7

Problems ............................................................... 249 Poisson Random Variables ....................................... 250

Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

CHAPTER 6

195 198 200 201

253 254 254 256

Normal Random Variables ........................................... 261 6.1 6.2

Introduction ........................................................... 262 Continuous Random Variables ................................. 262

6.3

Problems ............................................................... 264 Normal Random Variables ....................................... 266

6.4

Problems ............................................................... 269 Probabilities Associated with a Standard Normal Random Variable .................................................... 271

6.5

Problems ............................................................... 276 Finding Normal Probabilities: Conversion to the Standard Normal .................................................... 277

ix

x

Contents

6.6

Additive Property of Normal Random Variables .......... 279

6.7

Problems ............................................................... 281 Percentiles of Normal Random Variables ................... 284

Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

CHAPTER 7

Distributions of Sampling Statistics ............................. 297 7.1 7.2 7.3

A Preview.............................................................. 298 Introduction ........................................................... 298 Sample Mean ......................................................... 299

7.4

Problems ............................................................... 303 Central Limit Theorem ............................................ 304

7.5

7.4.1

Distribution of the Sample Mean.................... 306

7.4.2

How Large a Sample Is Needed? ................... 310

Problems ............................................................... 311 Sampling Proportions from a Finite Population ........... 313 7.5.1

7.6

Probabilities Associated with Sample Proportions: The Normal Approximation to the Binomial Distribution .............................. 317

Problems ............................................................... 319 Distribution of the Sample Variance of a Normal Population ............................................................. 323

Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

CHAPTER 8

289 290 290 293

325 325 326 327

Estimation ................................................................... 331 8.1 8.2

Introduction ........................................................... 332 Point Estimator of a Population Mean........................ 333

8.3

Problems ............................................................... 334 Point Estimator of a Population Proportion ................. 336 Problems ............................................................... 338 *8.3.1

8.4

Estimating the Probability of a Sensitive Event ......................................................... 341

Problems ............................................................... 342 Estimating a Population Variance ............................. 342 Problems ............................................................... 344

Contents

8.5

Interval Estimators of the Mean of a Normal Population with Known Population Variance .............. 347 8.5.1

8.6

8.6.1 8.7

Lower and Upper Conﬁdence Bounds ............ 355

Problems ............................................................... 357 Interval Estimators of the Mean of a Normal Population with Unknown Population Variance .......... 359 Lower and Upper Conﬁdence Bounds ............ 364

Problems ............................................................... 366 Interval Estimators of a Population Proportion ............ 371 8.7.1

Length of the Conﬁdence Interval .................. 373

8.7.2

Lower and Upper Conﬁdence Bounds ............ 375

Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

CHAPTER 9

377 380 381 383

Testing Statistical Hypotheses ..................................... 387 9.1 9.2

Introduction ........................................................... 388 Hypothesis Tests and Signiﬁcance Levels .................. 388

9.3

Problems ............................................................... 392 Tests Concerning the Mean of a Normal Population: Case of Known Variance.......................................... 394 Problems ............................................................... 400

9.4

9.5

9.3.1 One-Sided Tests .......................................... 403 The t Test for the Mean of a Normal Population: Case of Unknown Variance ...................................... 409 Problems ............................................................... 417 Hypothesis Tests Concerning Population Proportions . 421 9.5.1

Two-Sided Tests of p .................................... 425

Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems and Proposed Case Studies .....................

429 433 433 437

CHAPTER 10 Hypothesis Tests Concerning Two Populations ........... 443 10.1 Introduction ........................................................... 444 10.2 Testing Equality of Means of Two Normal Populations: Case of Known Variances...................... 446 Problems ............................................................... 450

xi

xii

Contents

10.3 Testing Equality of Means: Unknown Variances and Large Sample Sizes ................................................. 453 Problems ............................................................... 459 10.4 Testing Equality of Means: Small-Sample Tests when the Unknown Population Variances Are Equal... 463 Problems ............................................................... 468 10.5 Paired-Sample t Test ............................................... 471 Problems ............................................................... 476 10.6 Testing Equality of Population Proportions ................ 481 Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

490 493 493 498

CHAPTER 11 Analysis of Variance .................................................... 503 11.1 Introduction ........................................................... 504 11.2 One-Factor Analysis of Variance............................... 505 A Remark on the Degrees of Freedom ....................... 507 Problems ............................................................... 510 11.3 Two-Factor Analysis of Variance: Introduction and Parameter Estimation.............................................. 514 Problems ............................................................... 518 11.4 Two-Factor Analysis of Variance: Testing Hypotheses 520 Problems ............................................................... 11.5 Final Comments ..................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

527 529 530 530 533

CHAPTER 12 Linear Regression ........................................................ 537 12.1 Introduction ........................................................... 539 12.2 Simple Linear Regression Model ............................... 540 Problems ............................................................... 542 12.3 Estimating the Regression Parameters ...................... 544 Problems ............................................................... 548 12.4 Error Random Variable ............................................ 553 Problems ............................................................... 556 12.5 Testing the Hypothesis that β = 0 ............................ 557 Problems ............................................................... 560

Contents

12.6 Regression to the Mean ........................................... 564 *12.6.1 Why Biological Data Sets Are Often Normally Distributed.................................... 569 Problems ............................................................... 570 12.7 Prediction Intervals for Future Responses .................. 573 Problems ............................................................... 575 12.8 Coefﬁcient of Determination..................................... 578 Problems ............................................................... 580 12.9 Sample Correlation Coefﬁcient ................................. 582 Problems ............................................................... 583 12.10 Analysis of Residuals: Assessing the Model............... 584 Problems ............................................................... 586 12.11 Multiple Linear Regression Model............................. 586 12.11.1 Dummy Variables for Categorical Data ........... 590 Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

592 595 595 599

CHAPTER 13 Chi-Squared Goodness-of-Fit Tests .............................. 605 13.1 Introduction ........................................................... 606 13.2 Chi-Squared Goodness-of-Fit Tests ........................... 609 Problems ............................................................... 615 13.3 Testing for Independence in Populations Classiﬁed According to Two Characteristics ............................. 620 Problems ............................................................... 626 13.4 Testing for Independence in Contingency Tables with Fixed Marginal Totals ...................................... 631 Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

634 637 638 640

CHAPTER 14 Nonparametric Hypotheses Tests ................................ 647 14.1 Introduction ........................................................... 648 14.2 Sign Test ............................................................... 648 14.2.1 Testing the Equality of Population Distributions when Samples Are Paired.......... 652 14.2.2 One-Sided Tests .......................................... 653 Problems ............................................................... 655

xiii

xiv

Contents

14.3 Signed-Rank Test ................................................... 657 14.3.1 Zero Differences and Ties ............................. 662 Problems ............................................................... 664 14.4 Rank-Sum Test for Comparing Two Populations ......... 667 14.4.1 Comparing Nonparametric Tests with Tests that Assume Normal Distributions ................. 672 Problems ............................................................... 673 14.5 Runs Test for Randomness ...................................... 676 Problems ............................................................... 681 14.6 Testing the Equality of Multiple Probability Distributions .......................................................... 683 14.6.1 When the Data Are a Set of Comparison Rankings .................................................... 685 Problems ............................................................... 688 14.7 Permutation Tests .................................................. 689 Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

692 693 693 696

CHAPTER 15 Quality Control ............................................................ 699 15.1 Introduction ........................................................... 700 15.2 The X Control Chart for Detecting a Shift in the Mean . 700 Problems ............................................................... 705 15.2.1 When the Mean and Variance Are Unknown... 707 15.2.2 S Control Charts .......................................... 710 Problems ............................................................... 713 15.3 Control Charts for Fraction Defective ........................ 715 Problems ............................................................... 717 15.4 Exponentially Weighted Moving-Average Control Charts ................................................................... 717 Problems ............................................................... 721 15.5 Cumulative-Sum Control Charts ............................... 722 Problems ............................................................... Key Terms...................................................................... Summary ....................................................................... Review Problems ............................................................

725 725 725 726

Contents

APPENDICES

................................................................................ 727

APPENDIX A A Data Set ................................................................... 729 APPENDIX B Mathematical Preliminaries ......................................... 733 B.1 B.2 B.3

Summation ............................................................ 733 Absolute Value....................................................... 733 Set Notation ........................................................... 734

APPENDIX C How to Choose a Random Sample ............................... 735 APPENDIX D Tables ......................................................................... 739 Table D.1 Table D.2 Table D.3 Table D.4 Table D.5

Standard Normal Probabilities.......................... Percentiles tn,α of t Distributions ...................... 2 of the Chi-Squared Percentiles χn,α Distributions ................................................. Percentiles of F Distributions ........................... Binomial Distribution Function.........................

739 740 741 743 749

APPENDIX E Programs ..................................................................... 755 ANSWERS TO ODD-NUMBERED PROBLEMS ..................................... 757 INDEX .............................................................................................. 807

xv

This page intentionally left blank

Preface

Statistical thinking will one day be as necessary for efﬁcient citizenship as the ability to read and write. H. G. Wells (1866–1946)

In today’s complicated world, very few issues are clear-cut and without controversy. In order to understand and form an opinion about an issue, one must usually gather information, or data. To learn from data, one must know something about statistics, which is the art of learning from data. This introductory statistics text is written for college-level students in any ﬁeld of study. It can be used in a quarter, semester, or full-year course. Its only prerequisite is high school algebra. Our goal in writing it is to present statistical concepts and techniques in a manner that will teach students not only how and when to utilize the statistical procedures developed, but also to understand why these procedures should be used. As a result we have made a great effort to explain the ideas behind the statistical concepts and techniques presented. Concepts are motivated, illustrated, and explained in a way that attempts to increase one’s intuition. It is only when a student develops a feel or intuition for statistics that she or he is really on the path toward making sense of data. To illustrate the diverse applications of statistics and to offer students different perspectives about the use of statistics, we have provided a wide variety of text examples and problems to be worked by students. Most refer to real-world issues, such as gun control, stock price models, health issues, driving age limits, school admission ages, public policy issues, gender issues, use of helmets, sports, disputed authorship, scientiﬁc fraud, and Vitamin C, among many others. Many of them use data that not only are real but are themselves of interest. The examples have been posed in a clear and concise manner and include many thought-provoking problems that emphasize thinking and problem-solving skills. In addition, some of the problems are designed to be open-ended and can be used as starting points for term projects. xvii

xviii

Preface

SOME SPECIAL FEATURES OF THE TEXT Introduction The ﬁrst numbered section of each chapter is an introduction that poses a realistic statistical situation to help students gain perspective on what they will encounter in the chapter. Statistics in Perspective Statistics in Perspective highlights are placed throughout the book to illustrate real-world application of statistical techniques and concepts. These perspectives are designed to help students analyze and interpret data while utilizing proper statistical techniques and methodology. Real Data Throughout the text discussions, examples, perspective highlights, and problems, real data sets are used to enhance the students’ understanding of the material. These data sets provide information for the study of current issues in a variety of disciplines, such as health, medicine, sports, business, and education. Historical Perspectives These enrichment sections proﬁle prominent statisticians and historical events, giving students an understanding of how the discipline of statistics has evolved. Problems/Review Problems This text includes hundreds of exercises placed at the end of each section within a chapter, as well as more comprehensive review problems at the end of each chapter. Many of these problems utilize real data and are designed to assess the students’ conceptual as well as computational understanding of the material. Selected problems are open-ended and offer excellent opportunity for extended discussion, group activities, or student projects. Summary/Key Terms An end-of-chapter summary provides a detailed review of important concepts and formulas covered in the chapter. Key terms and their deﬁnitions are listed that serve as a working glossary within each chapter. Formula Summary Important tables and formulas that students often refer to and utilize are included on the inside front and back covers of the book. These can serve as a quick reference when doing homework or studying for an exam. Program CD-ROM A CD-ROM is provided with each volume that includes programs that can be used to solve basic statistical computation problems. Please refer to Appendix E for a listing of these programs.

THE TEXT In Chap. 1 we introduce the subject matter of statistics and present its two branches. The ﬁrst of these, called descriptive statistics, is concerned with the collection, description, and summarization of data. The second branch, called inferential statistics, deals with the drawing of conclusions from data.

Preface

Chapters 2 and 3 are concerned with descriptive statistics. In Chap. 2 we discuss tabular and graphical methods of presenting a set of data. We see that an effective presentation of a data set can often reveal certain of its essential features. Chap. 3 shows how to summarize certain features of a data set. In order to be able to draw conclusions from data it is necessary to have some understanding of what they represent. For instance, it is often assumed that the data constitute a “random sample from some population.” In order to understand exactly what this and similar phrases signify, it is necessary to have some understanding of probability, and that is the subject of Chap. 4. The study of probability is often a troublesome issue in an introductory statistics class because many students ﬁnd it a difﬁcult subject. As a result, certain textbooks have chosen to downplay its importance and present it in a rather cursory style. We have chosen a different approach and attempted to concentrate on its essential features and to present them in a clear and easily understood manner. Thus, we have brieﬂy but carefully dealt with the concept of the events of an experiment, the properties of the probabilities that are assigned to the events, and the idea of conditional probability and independence. Our study of probability is continued in Chap. 5, where discrete random variables are introduced, and in Chap. 6, which deals with the normal and other continuous random variables. Chapter 7 is concerned with the probability distributions of sampling statistics. In this chapter we learn why the normal distribution is of such importance in statistics. Chapter 8 deals with the problem of using data to estimate certain parameters of interest. For instance, we might want to estimate the proportion of people who are presently in favor of congressional term limits. Two types of estimators are studied. The ﬁrst of these estimates the quantity of interest with a single number (for instance, it might estimate that 52 percent of the voting population favors term limits). The second type provides an estimator in the form of an interval of values (for instance, it might estimate that between 49 and 55 percent of the voting population favors term limits). Chapter 9 introduces the important topic of statistical hypothesis testing, which is concerned with using data to test the plausibility of a speciﬁed hypothesis. For instance, such a test might reject the hypothesis that over 60 percent of the voting population favors term limits. The concept of p value, which measures the degree of plausibility of the hypothesis after the data have been observed, is introduced. Whereas the tests in Chap. 9 deal with a single population, the ones in Chap. 10 relate to two separate populations. For instance, we might be interested in testing whether the proportions of men and of women that favor term limits are the same.

xix

xx

Preface

Probably the most widely used statistical inference technique is that of the analysis of variance; this is introduced in Chap. 11. This technique allows us to test inferences about parameters that are affected by many different factors. Both oneand two-factor analysis of variance problems are considered in this chapter. In Chap. 12 we learn about linear regression and how it can be used to relate the value of one variable (say, the height of a man) to that of another (the height of his father). The concept of regression to the mean is discussed, and the regression fallacy is introduced and carefully explained. We also learn about the relation between regression and correlation. Also, in an optional section, we use regression to the mean along with the central limit theorem to present a simple, original argument to explain why biological data sets often appear to be normally distributed. In Chap. 13 we present goodness-of-ﬁt tests, which can be used to test whether a proposed model is consistent with data. This chapter also considers populations classiﬁed according to two characteristics and shows how to test whether the characteristics of a randomly chosen member of the population are independent. Chapter 14 deals with nonparametric hypothesis tests, which are tests that can be used in situations where the ones of earlier chapters are inappropriate. Chapter 15 introduces the subject matter of quality control, a key statistical technique in manufacturing and production processes.

NEW TO THIS EDITION The third edition has many new and updated examples and exercises. In addition, there is a new subsection (12.11.1) on the use of dummy variables in multiple regression models. There is also a new section (14.6) on the use of the KruskalWallis nonparametric test of the equality of multiple probability distributions, with a subsection (14.6.1) giving a discussion of the Freedman test which can be used to test this hypothesis when the data are comparison rankings. There is also a new section (14.7) on the class of nonparametric tests known as permutation tests.

Acknowledgments

We would like to thank the following reviewers of the third edition: Katherine T. Halvorsen, Smith College Liam O’Brien, Colby College In addition we wish to thank Margaret Lin, Erol Pekoz, and the following reviewers of the ﬁrst edition for their many helpful comments: William H. Beyer, University of Akron; Patricia Buchanan, Pennsylvania State University; Michael Eurgubian, Santa Rosa Junior College; Larry Griffey, Florida Community College, Jacksonville; James E. Holstein, University of Missouri; James Householder, Humboldt State University; Robert Lacher, South Dakota State University; Jacinta Mann, Seton Hill College; C. J. Park, San Diego State University; Ronald Pierce, Eastern Kentucky University; Lawrence Riddle, Agnes Scott College; Gaspard T. Rizzuto, University of Southwestern Louisiana; Jim Robison-Cox, Montana State University; Walter Rosenkrantz, University of Massachusetts, Amherst; Bruce Sisko, Belleville Area College; Glen Swindle, University of California, Santa Barbara; Paul Vetrano, Santa Rose Junior College; Joseph J. Walker, Georgia State University; Deborah White, College of the Redwoods; and Cathleen Zucco, LeMoyne College. Sheldon M. Ross

xxi

This page intentionally left blank

CHAPTER 1

Introduction to Statistics Statisticians have already overrun every branch of science with a rapidity of conquest rivalled only by Attila, Mohammed, and the Colorado beetle. Maurice Kendall (British statistician)

CONTENTS 1.1

Introduction .........................................................................

1

1.2

The Nature of Statistics .........................................................

3

1.3

Populations and Samples .......................................................

5

1.4

A Brief History of Statistics ....................................................

7

Key Terms ..................................................................................

10

The Changing Deﬁnition of Statistics .............................................

11

Review Problems .........................................................................

11

This chapter introduces the subject matter of statistics, the art of learning from data. It describes the two branches of statistics, descriptive and inferential. The idea of learning about a population by sampling and studying certain of its members is discussed. Some history is presented.

1.1 INTRODUCTION Is it better for children to start school at a younger or older age? This is certainly a question of interest to many parents as well as to people who set public policy. How can we answer it? It is reasonable to start by thinking about this question, relating it to your own experiences, and talking it over with friends. However, if you want to convince Introductory Statistics, DOI: 10.1016/B978-0-12-374388-6.00001-6 © 2010, Elsevier Inc. All rights reserved.

1

2

CH A P T E R 1: Introduction to Statistics

others and obtain a consensus, it is then necessary to gather some objective information. For instance, in many states, achievement tests are given to children at the end of their ﬁrst year in school. The children’s results on these tests can be obtained and then analyzed to see whether there appears to be a connection between children’s ages at school entrance and their scores on the test. In fact, such studies have been done, and they have generally concluded that older student entrants have, as a group, fared better than younger entrants. However, it has also been noted that the reason for this may just be that those students who entered at an older age would be older at the time of the examination, and this by itself may be what is responsible for their higher scores. For instance, suppose parents did not send their 6-year-olds to school but rather waited an additional year. Then, since these children will probably learn a great deal at home in that year, they will probably score higher when they take the test at the end of their ﬁrst year of school than they would have if they had started school at age 6. A recent study (Table 1.1) has attempted to improve upon earlier work by examining the effect of children’s age upon entering school on the eventual number of years of school completed. These authors argue that the total number of years spent in school is a better measure of school success than is a score on an achievement test taken in an early grade. Using 1960 and 1980 census data, they concluded that the age at which a child enters school has very little effect on the total number of years that a child spends in school. Table 1.1 is an abridgment of one presented in their work. The table indicates that for children beginning school in 1949, the younger half (whose average entrance age was 6.29 years) spent an average of 13.77 years, and the older half an average of 13.78 years, in school. Note that we have not presented the preceding in order to make the case that the ages at which children enter school do not affect their performance in school.

Table 1.1 Total Years in School Related to Starting Age Younger half of children

Older half of children

Year

Average age on starting school

Average number of years completed

Average age on starting school

Average number of years completed

1946 1947 1948 1949 1950 1951 1952

6.38 6.34 6.31 6.29 6.24 6.18 6.08

13.84 13.80 13.78 13.77 13.68 13.63 13.49

6.62 6.59 6.56 6.54 6.53 6.45 6.37

13.67 13.86 13.79 13.78 13.68 13.65 13.53

Source: J. Angrist and A. Krueger, “The Effect of Age at School Entry on Educational Attainment: An Application of Instrumental Variables with Moments from Two Samples,” Journal of the American Statistical Association, vol. 87, no. 18, 1992, pp. 328–336.

1.2 The Nature of Statistics

Rather we are using it to indicate the modern approach to learning about a complicated question. Namely, one must collect relevant information, or data, and these data must then be described and analyzed. Such is the subject matter of statistics.

1.2 THE NATURE OF STATISTICS It has become a truism in today’s world that in order to learn about something, you must ﬁrst collect data. For instance, the ﬁrst step in learning about such things as 1. 2. 3. 4. 5.

The present state of the economy The percentage of the voting public who favors a certain proposition The average miles per gallon of a newly developed automobile The efﬁcacy of a new drug The usefulness of a new way of teaching reading to children in elementary school

is to collect relevant data. Deﬁnition Statistics is the art of learning from data. It is concerned with the collection of data, their subsequent description, and their analysis, which often leads to the drawing of conclusions.

1.2.1 Data Collection Sometimes a statistical analysis begins with a given set of data; for instance, the government regularly collects and publicizes data about such quantities as the unemployment rate and the gross domestic product. Statistics would then be used to describe, summarize, and analyze these data. In other situations, data are not yet available, and statistics can be utilized to design an appropriate experiment to generate data. The experiment chosen should depend on the use that one wants to make of the data. For instance, if a cholesterol-lowering drug has just been developed and its efﬁcacy needs to be determined, volunteers will be recruited and their cholesterol levels noted. They will then be given the drug for some period, and their levels will be measured again. However, it would be an ineffective experiment if all the volunteers were given the drug. For if this were so, then even if the cholesterol levels of all the volunteers were signiﬁcantly reduced, we would not be justiﬁed in concluding that the improvements were due to the drug used and not to some other possibility. For instance, it is a well-documented fact that any medication received by a patient, whether or not it is directly related to that patient’s suffering, will often lead to an improvement in the patient’s condition. This is the placebo effect, which is not as surprising as it might seem at ﬁrst, since a patient’s belief that she or he is being effectively treated often leads to a reduction in stress, which can result in

3

4

CH A P T E R 1: Introduction to Statistics

an improved state of health. In addition, there might have been other—usually unknown—factors that played a role in the reduction of cholesterol levels. Perhaps the weather was unusually warm (or cold), causing the volunteers to spend more or less time outdoors than usual, and this was a factor. Thus, we see that the experiment that calls for giving the drug to all the volunteers is not well designed for generating data from which we can learn about the efﬁcacy of that drug. A better experiment is one that tries to neutralize all other possible causes of the change of cholesterol level except the drug. The accepted way of accomplishing this is to divide the volunteers into two groups; then one group receives the drug, and the other group receives a tablet (known as a placebo) that looks and tastes like the drug but has no physiological effect. The volunteers should not know whether they are receiving the true drug or the placebo, and indeed it is best if the medical people overseeing the experiment also do not know, so their own biases will not play a role. In addition, we want the division of the volunteers into the two groups to be done such that neither of the groups is favored in that it tends to have the “better” patients. The accepted best approach for arranging this is to break up the volunteers “at random,” where by this term we mean that the breakup is done in such a manner that all possible choices of people in the group receiving the drug are equally likely. The group that does not receive any treatment (that is, the volunteers that receive a placebo) is called the control group. At the end of the experiment, the data should be described. For instance, the before and after cholesterol levels of each volunteer should be presented, and the experimenter should note whether the volunteer received the drug or the placebo. In addition, summary measures such as the average reduction in cholesterol of members of the control group and members of the drug group should be determined. Deﬁnition The part of statistics concerned with the description and summarization of data is called descriptive statistics.

1.2.2 Inferential Statistics and Probability Models When the experiment is completed and the data are described and summarized, we hope to be able to draw a conclusion about the efﬁcacy of the drug. For instance, can we conclude that it is effective in reducing blood cholesterol levels? Deﬁnition The part of statistics concerned with the drawing of conclusions from data is called inferential statistics. To be able to draw a conclusion from the data, we must take into account the possibility of chance. For instance, suppose that the average reduction in cholesterol is lower for the group receiving the drug than for the control group. Can we conclude that this result is due to the drug? Or is it possible that the drug is really ineffective and that the improvement was just a chance occurrence? For instance,

1.3 Populations and Samples

the fact that a coin comes up heads 7 times in 10 ﬂips does not necessarily mean that the coin is more likely to come up heads than tails in future ﬂips. Indeed, it could be a perfectly ordinary coin that, by chance, just happened to land heads 7 times out of the total of 10 ﬂips. (On the other hand, if the coin had landed heads 47 times out of 50 ﬂips, then we would be quite certain that it was not an ordinary coin.) To be able to draw logical conclusions from data, it is usually necessary to make some assumptions about the chances (or probabilities) of obtaining the different data values. The totality of these assumptions is referred to as a probability model for the data. Sometimes the nature of the data suggests the form of the probability model that is assumed. For instance, suppose the data consist of the responses of a selected group of individuals to a question about whether they are in favor of a senator’s welfare reform proposal. Provided that this group was randomly selected, it is reasonable to suppose that each individual queried was in favor of the proposal with probability p, where p represents the unknown proportion of all citizens in favor of the proposal. The resultant data can then be used to make inferences about p. In other situations, the appropriate probability model for a given data set will not be readily apparent. However, a careful description and presentation of the data sometimes enable us to infer a reasonable model, which we can then try to verify with the use of additional data. Since the basis of statistical inference is the formulation of a probability model to describe the data, an understanding of statistical inference requires some knowledge of the theory of probability. In other words, statistical inference starts with the assumption that important aspects of the phenomenon under study can be described in terms of probabilities, and then it draws conclusions by using data to make inferences about these probabilities.

1.3 POPULATIONS AND SAMPLES In statistics, we are interested in obtaining information about a total collection of elements, which we will refer to as the population. The population is often too large for us to examine each of its members. For instance, we might have all the residents of a given state, or all the television sets produced in the last year by a particular manufacturer, or all the households in a given community. In such cases, we try to learn about the population by choosing and then examining a subgroup of its elements. This subgroup of a population is called a sample. Deﬁnition The total collection of all the elements that we are interested in is called a population. A subgroup of the population that will be studied in detail is called a sample.

5

6

CH A P T E R 1: Introduction to Statistics

In order for the sample to be informative about the total population, it must be, in some sense, representative of that population. For instance, suppose that we are interested in learning about the age distribution of people residing in a given city, and we obtain the ages of the ﬁrst 100 people to enter the town library. If the average age of these 100 people is 46.2 years, are we justiﬁed in concluding that this is approximately the average age of the entire population? Probably not, for we could certainly argue that the sample chosen in this case is not representative of the total population because usually more young students and senior citizens use the library than do working-age citizens. Note that representative does not mean that the age distribution of people in the sample is exactly that of the total population, but rather that the sample was chosen in such a way that all parts of the population had an equal chance to be included in the sample. In certain situations, such as the library illustration, we are presented with a sample and must then decide whether this sample is reasonably representative of the entire population. In practice, a given sample generally cannot be considered to be representative of a population unless that sample has been chosen in a random manner. This is because any speciﬁc nonrandom rule for selecting a sample often results in one that is inherently biased toward some data values as opposed to others. Deﬁnition A sample of k members of a population is said to be a random sample, sometimes called a simple random sample, if the members are chosen in such a way that all possible choices of the k members are equally likely. Thus, although it may seem paradoxical, we are most likely to obtain a representative sample by choosing its members in a totally random fashion without any prior considerations of the elements that will be chosen. In other words, we need not attempt to deliberately choose the sample so that it contains, for instance, the same gender percentage and the same percentage of people in each profession as found in the general population. Rather, we should just leave it up to “chance” to obtain roughly the correct percentages. The actual mechanics of choosing a random sample involve the use of random numbers and will be presented in App. C. Once a random sample is chosen, we can use statistical inference to draw conclusions about the entire population by studying the elements of the sample.

*1.3.1 Stratiﬁed Random Sampling A more sophisticated approach to sampling than simple random sampling is the stratiﬁed random sampling approach. This approach, which requires more initial information about the population than does simple random sampling, can be explained as follows. Consider a high school that contains 300 students in ∗ The

asterisk signiﬁes optional material not used in the sequel.

1.4 A Brief History of Statistics

the ﬁrst-year class, 500 in the second-year class, and 600 each in the third- and fourth-year classes. Suppose that in order to learn about the students’ feelings concerning a military draft for 18-year-olds, an in-depth interview of 100 students will be done. Rather than randomly choosing 100 people from the 2000 students, in a stratiﬁed sample one calculates how many to choose from each class. Since the proportion of students who are ﬁrst-year is 300/2000 = 0.15, in a stratiﬁed sample the percentage is the same and thus there are 100 × 0.15 = 15 ﬁrst-year students in the sample. Similarly, one selects 100 × 0.25 = 25 secondyear students and 100 × 0.30 = 30 third-year and 30 fourth-year students. Then one selects students from each class at random. In other words, in this type of sample, ﬁrst the population is stratiﬁed into subpopulations, and then the correct number of elements is randomly chosen from each of the subpopulations. As a result, the proportions of the sample members that belong to each of the subpopulations are exactly the same as the proportions for the total population. Stratiﬁcation is particularly effective for learning about the “average” member of the entire population when there are inherent differences between the subpopulations with respect to the question of interest. For instance, in the foregoing survey, the upper-grade students, being older, would be more immediately affected by a military draft than the lower-grade students. Thus, each class might have inherently different feelings about the draft, and stratiﬁcation would be effective in learning about the feelings of the average student.

1.4 A BRIEF HISTORY OF STATISTICS A systematic collection of data on the population and the economy was begun in the Italian city-states of Venice and Florence during the Renaissance. The term statistics, derived from the word state, was used to refer to a collection of facts of interest to the state. The idea of collecting data spread from Italy to the other countries of western Europe. Indeed, by the ﬁrst half of the 16th century, it was common for European governments to require parishes to register births, marriages, and deaths. Because of poor public health conditions this last statistic was of particular interest. The high mortality rate in Europe before the 19th century was due mainly to epidemic diseases, wars, and famines. Among epidemics the worst were the plagues. Starting with the Black Plague in 1348, plagues recurred frequently for nearly 400 years. In 1562, as a way to alert the King’s court to consider moving to the countryside, the city of London began to publish weekly bills of mortality. Initially these mortality bills listed the places of death and whether a death had resulted from plague. Beginning in 1625, the bills were expanded to include all causes of death. In 1662 the English tradesman John Graunt published a book entitled Natural and Political Observations Made upon the Bills of Mortality. Table 1.2, which notes

7

8

CH A P T E R 1: Introduction to Statistics

Table 1.2 Total Deaths in England Year

Burials

Plague deaths

1592 1593 1603 1625 1636

25,886 17,844 37,294 51,758 23,359

11,503 10,662 30,561 35,417 10,400

the total number of deaths in England and the number due to the plague for ﬁve different plague years, is taken from this book. Graunt used the London bills of mortality to estimate the city’s population. For instance, to estimate the population of London in 1660, Graunt surveyed households in certain London parishes (or neighborhoods) and discovered that, on average, there were approximately 3 deaths for every 88 people. Dividing by 3 shows that, on average, there was roughly 1 death for every 88/3 people. Since the London bills cited 13,200 deaths in London for that year, Graunt estimated the London population to be about 13,200 ·

88 = 387,200 3

Graunt used this estimate to project a ﬁgure for all England. In his book he noted that these ﬁgures would be of interest to the rulers of the country, as indicators of both the number of men who could be drafted into an army and the number who could be taxed. Graunt also used the London bills of mortality—and some intelligent guesswork as to what diseases killed whom and at what age—to infer ages at death. (Recall that the bills of mortality listed only causes and places of death, not the ages of those dying.) Graunt then used this information to compute tables giving the proportion of the population that dies at various ages. Table 1.3 is one of Graunt’s mortality tables. It states, for instance, that of 100 births, 36 people will die before reaching age 6, 24 will die between the ages of 6 and 15, and so on. Graunt’s estimates of the ages at which people were dying were of great interest to those in the business of selling annuities. Annuities are the opposite of life insurance, in that one pays in a lump sum as an investment and then receives regular payments for as long as one lives. Graunt’s work on mortality tables inspired further work by Edmund Halley in 1693. Halley, the discoverer of the comet bearing his name (and also the man who was most responsible, by both his encouragement and his ﬁnancial support, for the publication of Isaac Newton’s famous Principia Mathematica), used tables

1.4 A Brief History of Statistics

Table 1.3 Graunt’s Mortality Table Age at death 0–6 6–16 16–26 26–36 36–46 46–56 56–66 66–76 ≥76

Deaths per 100 births 36 24 15 9 6 4 3 2 1

Note: The categories go up to, but do not include, the right-hand value. For instance, 0–6 means ages 0 through 5 years.

of mortality to compute the odds that a person of any age would live to any other particular age. Halley was inﬂuential in convincing the insurers of the time that an annual life insurance premium should depend on the age of the person being insured. Following Graunt and Halley, the collection of data steadily increased throughout the remainder of the 17th century and on into the 18th century. For instance, the city of Paris began collecting bills of mortality in 1667; and by 1730 it had become common practice throughout Europe to record ages at death. The term statistics, which was used until the 18th century as a shorthand for the descriptive science of states, in the 19th century became increasingly identiﬁed with numbers. By the 1830s the term was almost universally regarded in Britain and France as being synonymous with the numerical science of society. This change in meaning was caused by the large availability of census records and other tabulations that began to be systematically collected and published by the governments of western Europe and the United States beginning around 1800. Throughout the 19th century, although probability theory had been developed by such mathematicians as Jacob Bernoulli, Karl Friedrich Gauss, and Pierre Simon Laplace, its use in studying statistical ﬁndings was almost nonexistent, as most social statisticians at the time were content to let the data speak for themselves. In particular, at that time statisticians were not interested in drawing inferences about individuals, but rather were concerned with the society as a whole. Thus, they were not concerned with sampling but rather tried to obtain censuses of the entire population. As a result, probabilistic inference from samples to a population was almost unknown in 19th-century social statistics. It was not until the late 1800s that statistics became concerned with inferring conclusions from numerical data. The movement began with Francis Galton’s

9

10

CH A P T E R 1: Introduction to Statistics

work on analyzing hereditary genius through the uses of what we would now call regression and correlation analysis (see Chap. 12) and obtained much of its impetus from the work of Karl Pearson. Pearson, who developed the chi-squared goodness-of-ﬁt test (see Chap. 13), was the ﬁrst director of the Galton laboratory, endowed by Francis Galton in 1904. There Pearson originated a research program aimed at developing new methods of using statistics in inference. His laboratory invited advanced students from science and industry to learn statistical methods that could then be applied in their ﬁelds. One of his earliest visiting researchers was W. S. Gosset, a chemist by training, who showed his devotion to Pearson by publishing his own works under the name Student. (A famous story has it that Gosset was afraid to publish under his own name for fear that his employers, the Guinness brewery, would be unhappy to discover that one of its chemists was doing research in statistics.) Gosset is famous for his development of the t test (see Chap. 9). Two of the most important areas of applied statistics in the early 20th century were population biology and agriculture. This was due to the interest of Pearson and others at his laboratory and to the remarkable accomplishments of the English scientist Ronald A. Fisher. The theory of inference developed by these pioneers, including, among others, Karl Pearson’s son Egon and the Polish-born mathematical statistician Jerzy Neyman, was general enough to deal with a wide range of quantitative and practical problems. As a result, after the early years of this century, a rapidly increasing number of people in science, business, and government began to regard statistics as a tool able to provide quantitative solutions to scientiﬁc and practical problems. Nowadays the ideas of statistics are everywhere. Descriptive statistics are featured in every newspaper and magazine. Statistical inference has become indispensable to public health and medical research, to marketing and quality control, to education, to accounting, to economics, to meteorological forecasting, to polling and surveys, to sports, to insurance, to gambling, and to all research that makes any claim to being scientiﬁc. Statistics has indeed become ingrained in our intellectual heritage.

KEY TERMS Statistics: The art of learning from data. Descriptive statistics: The part of statistics that deals with the description and summarization of data. Inferential statistics: The part of statistics that is concerned with drawing conclusions from data. Probability model: The mathematical assumptions relating to the likelihood of different data values.

Review Problems

Population: A collection of elements of interest. Sample: A subgroup of the population that is to be studied. Random sample of size k: A sample chosen in such a manner that all subgroups of size k are equally likely to be selected. Stratiﬁed random sample: A sample obtained by dividing the population into distinct subpopulations and then choosing random samples from each subpopulation.

THE CHANGING DEFINITION OF STATISTICS Statistics has then for its object that of presenting a faithful representation of a state at a determined epoch. (Quetelet, 1849) Statistics are the only tools by which an opening can be cut through the formidable thicket of difﬁculties that bars the path of those who pursue the Science of man. (Galton, 1889) Statistics may be regarded (i) as the study of populations, (ii) as the study of variation, and (iii) as the study of methods of the reduction of data. (Fisher, 1925) Statistics is a scientiﬁc discipline concerned with collection, analysis, and interpretation of data obtained from observation or experiment. The subject has a coherent structure based on the theory of Probability and includes many different procedures which contribute to research and development throughout the whole of Science and Technology. (E. Pearson, 1936) Statistics is the name for that science and art which deals with uncertain inferences—which uses numbers to ﬁnd out something about nature and experience. (Weaver, 1952) Statistics has become known in the 20th century as the mathematical tool for analyzing experimental and observational data. (Porter, 1986) Statistics is the art of learning from data. (Ross, 2010)

REVIEW PROBLEMS 1. This problem refers to Table 1.1. (a) In which year was there the largest difference between the average number of years of school completed by the younger and older starters? (b) Were there more years in which the average number of years completed by the younger starting group exceeded that of the older group, or the opposite?

11

12

CH A P T E R 1: Introduction to Statistics

2. The following is a graph of milk product consumption in the United States from 1909 to 2000. What general conclusion would you draw?

3. The following data yield the percentages of U.S. adults, characterized by educational level, that smoked in the years from 1999 to 2002. (a) For which group has there been a steady decline? (b) Would you say there is an overall trend? Cigarette Use in the U.S. (% of all adults) 1999 2000 2001 2002 Total Sex Male Female Education Non-high school graduate High school graduate Some college College graduate

25.8

24.9

24.9

26.0

28.3 23.4

26.9 23.1

27.1 23.0

28.7 23.4

39.9 36.4 32.5 18.2

32.4 31.1 27.7 13.9

33.8 32.1 26.7 13.8

35.2 32.3 29.0 14.5

4. A medical researcher, trying to establish the efﬁcacy of a new drug, has begun testing the drug along with a placebo. To make sure that the two groups of volunteer patients—those receiving the drug and those receiving a placebo—are as nearly alike as possible, the researcher has decided not to rely on chance but rather to carefully scrutinize the volunteers and then choose the groupings himself. Is this approach advisable? Why or why not? 5. Explain why it is important that a researcher who is trying to learn about the usefulness of a new drug not know which patients are receiving the new drug and which are receiving a placebo. 6. An election will be held next week, and by polling a sample of the voting population we are trying to predict whether the Republican or Democratic candidate will prevail. Which of the following methods of selection will yield a representative sample?

Review Problems

7.

8.

9. 10.

11.

(a) Poll all people of voting age attending a college basketball game. (b) Poll all people of voting age leaving a fancy midtown restaurant. (c) Obtain a copy of the voter registration list, randomly choose 100 names, and question them. (d) Use the results of a television call-in poll, in which the station asked its viewers to call and tell their choice. (e) Choose names from the telephone directory and call these people. The approach used in Prob. 6e led to a disastrous prediction in the 1936 Presidential election, in which Franklin Roosevelt defeated Alfred Landon by a landslide. A Landon victory had been predicted by the Literary Digest. The magazine based its prediction on the preferences of a sample of voters chosen from lists of automobile and telephone owners. (a) Why do you think the Literary Digest’s prediction was so far off? (b) Has anything changed between 1936 and now that would make you believe that the approach used by the Literary Digest would work better today? A researcher is trying to discover the average age at death for people in the United States today. To obtain data, the obituary columns of The New York Times are read for 30 days, and the ages at death of people in the United States are noted. Do you think this approach will lead to a representative sample? If, in Prob. 8, the average age at death of those recorded is 82.4 years, what conclusion could you draw? To determine the proportion of people in your town who are smokers, it has been decided to poll people at one of the following local spots: (a) The pool hall (b) The bowling alley (c) The shopping mall (d) The library Which of these potential polling places would most likely result in a reasonable approximation to the desired proportion? Why? A university plans on conducting a survey of its recent graduates to determine information on their yearly salaries. It randomly selected 200 recent graduates and sent them questionnaires dealing with their present jobs. Of these 200, however, only 86 questionnaires were returned. Suppose that the average of the yearly salaries reported was $75,000. (a) Would the university be correct in thinking that $75,000 was a good approximation to the average salary level of all its graduates? Explain the reasoning behind your answer.

13

14

CH A P T E R 1: Introduction to Statistics

12.

13. 14.

15.

16. 17.

18. 19.

(b) If your answer to (a) is no, can you think of any set of conditions relating to the group that returns questionnaires for which $75,000 would be a good approximation? An article reported that a survey of clothing worn by pedestrians killed at night in trafﬁc accidents revealed that about 80 percent of the victims were wearing dark-colored clothing and 20 percent were wearing light-colored clothing. The conclusion drawn in the article was that it is safer to wear light-colored clothing at night. (a) Is this conclusion justiﬁed? Explain. (b) If your answer to (a) is no, what other information would be needed before a ﬁnal conclusion could be drawn? Critique Graunt’s method for estimating the population of London. What implicit assumption is he making? The London bills of mortality listed 12,246 deaths in 1658. Supposing that a survey of London parishes showed that roughly 2 percent of the population died that year, use Graunt’s method to estimate London’s population in 1658. Suppose you were a seller of annuities in 1662, when Graunt’s book was published. Explain how you would make use of his data on the ages at which people were dying. Based on Table 1.2, which of the ﬁve plague years appears to have been the most severe? Explain your reasoning. Based on Graunt’s mortality table: (a) What proportion of babies survived to age 6? (b) What proportion survived to age 46? (c) What proportion died between the ages of 6 and 36? Why do you think that the study of statistics is important in your ﬁeld? How do you expect to utilize it in your future work? The chart on the following page gives the demographic and socioeconomic characteristics of adult smokers in upstate New York in 2006. Use it to determine if the following statements appear to be true. Answer yes or no. (a) A higher proportion of men than of women are current smokers. (b) The longer a person has been out of work, the more likely that person is a smoker. (c) The more education a person has, the more likely that person is to smoke. (d) Ethnicity does not appear to be related to smoking prevalence. It should be noted that even when the answer to a preceding question is yes that does not necessarily mean that the characteristic is a cause of smoking, but only that there is a positive association between it and smoking. The concept of association, or correlation, will be considered in Chap. 3.

Review Problems

15

This page intentionally left blank

CHAPTER 2

Describing Data Sets Numbers constitute the only universal language. Nathaniel West

People who don’t count won’t count. Anatole France

CONTENTS 2.1

Introduction .........................................................................

18

2.2

Frequency Tables and Graphs................................................

18

Problems .............................................................................

25

Grouped Data and Histograms ...............................................

32

Problems .............................................................................

39

Stem-and-Leaf Plots ..............................................................

44

Problems .............................................................................

47

Sets of Paired Data................................................................

51

Problems .............................................................................

54

Some Historical Comments ....................................................

58

2.3 2.4 2.5 2.6

Key Terms ..................................................................................

59

Summary ....................................................................................

60

Review Problems .........................................................................

63

In this chapter we learn methods for presenting and describing sets of data. We introduce different types of tables and graphs, which enable us to easily see key features of a data set.

Introductory Statistics, DOI: 10.1016/B978-0-12-374388-6.00002-8 © 2010, Elsevier Inc. All rights reserved.

17

18

CH A P T E R 2: Describing Data Sets

2.1 INTRODUCTION It is very important that the numerical ﬁndings of any study be presented clearly and concisely and in a manner that enables one to quickly obtain a feel for the essential characteristics of the data. This is particularly needed when the set of data is large, as is frequently the case in surveys or controlled experiments. Indeed, an effective presentation of the data often quickly reveals important features such as their range, degree of symmetry, how concentrated or spread out they are, where they are concentrated, and so on. In this chapter we will be concerned with techniques, both tabular and graphic, for presenting data sets. Frequency tables and frequency graphs are presented in Sec. 2.2. These include a variety of tables and graphs—line graphs, bar graphs, and polygon graphs—that are useful for describing data sets having a relatively small number of distinct values. As the number of distinct values becomes too large for these forms to be effective, it is useful to break up the data into disjoint classes and consider the number of data values that fall in each class. This is done in Sec. 2.3, where we study the histogram, a bar graph that results from graphing class frequencies. A variation of the histogram, called a stem-and-leaf plot, which uses the actual data values to represent the size of a class, is studied in Sec. 2.4. In Sec. 2.5 we consider the situation where the data consist of paired values, such as the population and the crime rate of various cities, and introduce the scatter diagram as an effective way of presenting such data. Some historical comments are presented in Sec. 2.6.

2.2 FREQUENCY TABLES AND GRAPHS The following data represent the number of days of sick leave taken by each of 50 workers of a given company over the last 6 weeks: 2, 2, 0, 0, 5, 8, 3, 4, 1, 0, 0, 7, 1, 7, 1, 5, 4, 0, 4, 0, 1, 8, 9, 7, 0, 1, 7, 2, 5, 5, 4, 3, 3, 0, 0, 2, 5, 1, 3, 0, 1, 0, 2, 4, 5, 0, 5, 7, 5, 1 Since this data set contains only a relatively small number of distinct, or different, values, it is convenient to represent it in a frequency table, which presents each distinct value along with its frequency of occurrence. Table 2.1 is a frequency table of the preceding data. In Table 2.1 the frequency column represents the number of occurrences of each distinct value in the data set. Note that the sum of all the frequencies is 50, the total number of data observations.

■

Example 2.1 Use Table 2.1 to answer the following questions: (a) How many workers had at least 1 day of sick leave?

2.2 Frequency Tables and Graphs

Table 2.1 A Frequency Table of Sick Leave Data Value 0 1 2 3 4

Frequency

Value

Frequency

12 8 5 4 5

5 6 7 8 9

8 0 5 2 1

(b) How many workers had between 3 and 5 days of sick leave? (c) How many workers had more than 5 days of sick leave? Solution (a) Since 12 of the 50 workers had no days of sick leave, the answer is 50 − 12 = 38. (b) The answer is the sum of the frequencies for values 3, 4, and 5; that is, 4 + 5 + 8 = 17. (c) The answer is the sum of the frequencies for the values 6, 7, 8, and 9. Therefore, the answer is 0 + 5 + 2 + 1 = 8. ■

2.2.1 Line Graphs, Bar Graphs, and Frequency Polygons Data from a frequency table can be graphically pictured by a line graph, which plots the successive values on the horizontal axis and indicates the corresponding frequency by the height of a vertical line. A line graph for the data of Table 2.1 is shown in Fig. 2.1.

FIGURE 2.1 A line graph.

19

20

CH A P T E R 2: Describing Data Sets

Sometimes the frequencies are represented not by lines but rather by bars having some thickness. These graphs, called bar graphs, are often utilized. Figure 2.2 presents a bar graph for the data of Table 2.1. Another type of graph used to represent a frequency table is the frequency polygon, which plots the frequencies of the different data values and then connects the plotted points with straight lines. Figure 2.3 presents the frequency polygon of the data of Table 2.1. A set of data is said to be symmetric about the value x0 if the frequencies of the values x0 − c and x0 + c are the same for all c. That is, for every constant c, there

FIGURE 2.2 A bar graph.

FIGURE 2.3 A frequency polygon.

2.2 Frequency Tables and Graphs

Table 2.2 Frequency Table of a Symmetric Data Set Value 0 2 3

Frequency

Value

Frequency

1 2 3

4 6 0

2 1 0

FIGURE 2.4 Bar graphs and symmetry.

are just as many data points that are c less than x0 as there are that are c greater than x0 . The data set presented in Table 2.2, a frequency table, is symmetric about the value x0 = 3. Data that are “close to” being symmetric are said to be approximately symmetric. The easiest way to determine whether a data set is approximately symmetric is to represent it graphically. Figure 2.4 presents three bar graphs: one of a symmetric data set, one of an approximately symmetric data set, and one of a data set that exhibits no symmetry.

2.2.2 Relative Frequency Graphs It is sometimes convenient to consider and plot the relative rather than the absolute frequencies of the data values. If f represents the frequency of occurrence of some data value x, then the relative frequency f/n can be plotted versus x, where n represents the total number of observations in the data set. For the data of Table 2.1, n = 50 and so the relative frequencies are as given in Table 2.3. Note that whereas the sum of the frequency column should be the total number of observations in the data set, the sum of the relative frequency column should be 1. A polygon plot of these relative frequencies is presented in Fig. 2.5. A plot of the relative frequencies looks exactly like a plot of the absolute frequencies, except that the labels on the vertical axis are the old labels divided by the total number of observations in the data set.

21

22

CH A P T E R 2: Describing Data Sets

To Construct a Relative Frequency Table from a Data Set Arrange the data set in increasing order of values. Determine the distinct values and how often they occur. List these distinct values alongside their frequencies f and their relative frequencies f/n, where n is the total number of observations in the data set. Table 2.3 Relative Frequencies, n = 50, of Sick Leave Data Value x

Frequency f

0

12

1

8

2

5

3

4

4

5

5

8

6

0

7

5

8

2

9

1

FIGURE 2.5 A relative frequency polygon.

Relative frequency f/n 12 50 8 50 5 50 4 50 5 50 8 50 0 50 5 50 2 50 1 50

= 0.24 = 0.16 = 0.10 = 0.08 = 0.10 = 0.16 = 0.00 = 0.10 = 0.04 = 0.02

2.2 Frequency Tables and Graphs

■

Example 2.2 The Masters Golf Tournament is played each year at the Augusta National Golf Club in Augusta, Georgia. To discover what type of score it takes to win this tournament, we have gathered all the winning scores from 1968 to 2004. The Masters Golf Tournament Winners Score

Year

Winner

Bob Goalby

277

1987

Larry Mize

285

George Archer

281

1988

Sandy Lyle

281

1970

Billy Casper

279

1989

Nick Faldo

283

1971

Charles Coody

279

1990

Nick Faldo

278

1972

Jack Nicklaus

286

1991

Ian Woosnam

277

1973

Tommy Aaron

283

1992

Fred Couples

275

1974

Gary Player

278

1993

Bernhard Langer

277

1975

Jack Nicklaus

276

1994

J.M. Olazabal

279

1976

Ray Floyd

271

1995

Ben Crenshaw

274

1977

Tom Watson

276

1996

Nick Faldo

276

1978

Gary Player

277

1997

Tiger Woods

270

1979

Fuzzy Zoeller

280

1998

Mark O’Meara

279

1980

Severiano Ballesteros

275

1999

J.M. Olazabal

280

1981

Tom Watson

280

2000

Vijay Singh

278

1982

Craig Stadler

284

2001

Tiger Woods

272

1983

Severiano Ballesteros

280

2002

Tiger Woods

276

Year

Winner

1968 1969

Score

1984

Ben Crenshaw

277

2003

Mike Weir

281

1985

Bernhard Langer

282

2004

Phil Mickelson

279

1986

Jack Nicklaus

279

(a) Arrange the data set of winning scores in a relative frequency table. (b) Plot these data in a relative frequency bar graph. Solution (a) The 37 winning scores range from a low of 270 to a high of 289. This is the relative frequency table: Winning score 270 271 272 274 275

Frequency f

Relative frequency f/37

1 1 1 1 2

0.027 0.027 0.027 0.027 0.054

23

24

CH A P T E R 2: Describing Data Sets

Winning score 276 277 278 279 280 281 282 283 284 285 286

Frequency f

Relative Frequency f/37

4 5 3 6 4 3 1 2 1 1 1

0.108 0.135 0.081 0.162 0.108 0.081 0.027 0.054 0.027 0.027 0.027

(b) The following is a relative frequency bar graph of the preceding data.

■

2.2.3 Pie Charts A pie chart is often used to plot relative frequencies when the data are nonnumeric. A circle is constructed and then is sliced up into distinct sectors, one for each different data value. The area of each sector, which is meant to represent the relative frequency of the value that the sector represents, is determined as follows. If the relative frequency of the data value is f/n, then the area of the sector is the fraction f/n of the total area of the circle. For instance, the data in Table 2.4 give the

2.2 Frequency Tables and Graphs

Table 2.4 Murder Weapons Type of weapon

Percentage of murders caused by this weapon

Handgun Knife Shotgun Rifle Personal weapon Other

52 18 7 4 6 13

FIGURE 2.6 A pie chart.

relative frequencies of types of weapons used in murders in a large midwestern city in 1985. These data are represented in a pie chart in Fig. 2.6. If a data value has relative frequency f/n, then its sector can be obtained by setting the angle at which the lines of the sector meet equal to 360 f/n degrees. For instance, in Fig. 2.6, the angle of the lines forming the knife sector is 360(0.18) = 64.8◦ .

PROBLEMS 1. The following data represent the sizes of 30 families that reside in a small town in Guatemala: 5, 13, 9, 12, 7, 4, 8, 6, 6, 10, 7, 11, 10, 8, 15, 8, 6, 9, 12, 10, 7, 11, 10, 8, 12, 9, 7, 10, 7, 8

25

26

CH A P T E R 2: Describing Data Sets

(a) Construct a frequency table for these data. (b) Using a line graph, plot the data. (c) Plot the data as a frequency polygon. 2. The following frequency table relates the weekly sales of bicycles at a given store over a 42-week period. Value

0

1

2

3

4

5

6

7

Frequency

3

6

7

10

8

5

2

1

(a) In how many weeks were at least 2 bikes sold? (b) In how many weeks were at least 5 bikes sold? (c) In how many weeks were an even number of bikes sold? 3. Fifteen fourth-graders were asked how many blocks they lived from school. The results are displayed in the following graph.

(a) What is the maximum number of blocks any student lives from school? (b) What is the minimum number of blocks? (c) How many students live less than 5 blocks from school? (d) How many students live more than 4 blocks from school? 4. Label each of the following data sets as symmetric, approximately symmetric, or not at all symmetric. A: 6, 0, 2, 1, 8, 3, 5 B: 4, 0, 4, 0, 2, 1, 3, 2 C: 1, 1, 0, 1, 0, 3, 3, 2, 2, 2 D: 9, 9, 1, 2, 3, 9, 8, 4, 5 5. The following table lists all the values but only some of the frequencies for a symmetric data set. Fill in the missing numbers. Value Frequency 10 20 30 40 50 60

8 7 3

2.2 Frequency Tables and Graphs

6. The following are the scores of 32 students who took a statistics test: 55, 70, 80, 75, 90, 80, 60, 100, 95, 70, 75, 85, 80, 80, 70, 95, 100, 80, 85, 70, 85, 90, 80, 75, 85, 70, 90, 60, 80, 70, 85, 80 Represent this data set in a frequency table, and then draw a bar graph. 7. Draw a relative frequency table for the data of Prob. 1. Plot these relative frequencies in a line graph. 8. The following data represent the time to tumor progression, measured in months, for 65 patients having a particular type of brain tumor called glioblastoma: 6, 5, 37, 10, 22, 9, 2, 16, 3, 3, 11, 9, 5, 14, 11, 3, 1, 4, 6, 2, 7, 3, 7, 5, 4, 8, 2, 7, 13, 16, 15, 9, 4, 4, 2, 3, 9, 5, 11, 3, 7, 5, 9, 3, 8, 9, 4, 10, 3, 2, 7, 6, 9, 3, 5, 4, 6, 4, 14, 3, 12, 6, 8, 12, 7 (a) Make up a relative frequency table for this data set. (b) Plot the relative frequencies in a frequency polygon. (c) Is this data set approximately symmetric? 9. The following relative frequency table is obtained from a data set of the number of emergency appendectomies performed each month at a certain hospital. Value Relative frequency

0

1

2

3

4

5

6

7

0.05

0.08

0.12

0.14

0.16

0.20

0.15

0.10

(a) What proportion of months has fewer than 2 emergency appendectomies? (b) What proportion of months has more than 5? (c) Is this data set symmetric? 10. Relative frequency tables and plots are particularly useful when we want to compare different sets of data. The following two data sets relate the number of months from diagnosis to death of AIDS patients for samples of male and female AIDS sufferers in the early years of the epidemic. Males

15

13

16

10

8

20

14

19

9

12

16

18

20

12

14

Females

8

12

10

8

14

12

13

11

9

8

9

10

14

9

10

14

27

28

CH A P T E R 2: Describing Data Sets

Plot these two data sets together in a relative frequency polygon. Use a different color for each set. What conclusion can you draw about which data set tends to have larger values? 11. Using the data of Example 2.2, determine the proportion of winning scores in the Masters Golf Tournament that is (a) Below 280 (b) 282 or higher (c) Between 278 and 284 inclusive The table on the following three pages gives the average number of days in each month that various cities have at least 0.01 inch of precipitation. Problems 12 through 14 refer to it. 12. Construct a relative frequency table for the average number of rainy days in January for the different cities. Then plot the data in a relative frequency polygon. 13. Using only the data relating to the ﬁrst 12 cities listed, construct a frequency table for the average number of rainy days in either November or December. 14. Using only the data relating to the ﬁrst 24 cities, construct relative frequency tables for the month of June and separately for the month of December. Then plot these two sets of data together in a relative frequency polygon. 15. The following table gives the number of deaths on British roads in 1987 for individuals in various classiﬁcations. Classiﬁcation Pedestrians Bicyclists Motorcyclists Automobile drivers

Number of deaths 1699 280 650 1327

Express this data set in a pie chart. 16. The following data, taken from The New York Times, represent the percentage of items, by total weight, in the garbage of New York City. Represent them in a pie chart. Organic material (food, yard waste, lumber, etc.) 37.3 Paper 30.8 Bulk (furniture, refrigerators, etc.) 10.9 Plastic 8.5 Glass 5 Metal 4 Inorganic 2.2 Aluminum 0.9 Hazardous waste 0.4

Average Number of Days with Precipitation of 0.01 Inch or More State

City

AL

Mobile

AK

Juneau

Length of record (yr.)

Jan.

Feb.

Mar.

Apr.

May

June

July

Aug.

Sept.

Oct.

Nov.

Dec.

Annual

46

11

10

11

7

8

11

16

14

10

6

8

10

123

43

18

17

18

17

17

16

17

18

20

24

19

21

220

AZ

Phoenix

48

4

4

4

2

1

1

4

5

3

3

3

4

36

AR

Little Rock

45

9

9

10

10

10

8

8

7

7

7

8

9

103

CA

Los Angeles

52

6

6

6

3

1

1

1

0

1

2

4

5

36

Sacramento

48

10

9

9

5

3

1

0

0

1

3

7

9

58

San Diego

47

7

6

7

5

2

1

0

1

1

3

5

6

43

San Francisco

60

11

10

10

6

3

1

0

0

1

4

7

10

62

CO

Denver

53

6

6

9

9

11

9

9

9

6

5

5

5

89

CT

Hartford

33

11

10

11

11

12

11

10

10

9

8

11

12

127

DE

Wilmington

40

11

10

11

11

11

10

9

9

8

8

10

10

117

DC

Washington

46

10

9

11

10

11

10

10

9

8

7

8

9

111

FL

Jacksonville

46

8

8

8

6

8

12

15

14

13

9

6

8

116

Miami

45

6

6

6

6

10

15

16

17

17

14

9

7

129

GA

Atlanta

53

11

10

11

9

9

10

12

9

8

6

8

10

115

HI

Honolulu

38

10

9

9

9

7

6

8

6

7

9

9

10

100

ID

Boise

48

12

10

10

8

8

6

2

3

4

6

10

11

91

IL

Chicago

29

11

10

12

12

11

10

10

9

10

9

10

12

127

Peoria

48

9

8

11

12

11

10

9

8

9

8

9

10

114

IN

Indianapolis

48

12

10

13

12

12

10

9

9

8

8

10

12

125

IA

Des Moines

48

7

7

10

11

11

11

9

9

9

8

7

8

107

KS

Wichita

34

6

5

8

8

11

9

7

8

8

6

5

6

86

KY

Louisville

40

11

11

13

12

12

10

11

8

8

8

10

11

125

LA

New Orleans

39

10

9

9

7

8

11

15

13

10

6

7

10

114

ME

Portland

47

11

10

11

12

13

11

10

9

8

9

12

12

128

MD

Baltimore

37

10

9

11

11

11

9

9

10

7

7

9

9

113

MA

Boston

36

12

10

12

11

12

11

9

10

9

9

11

12

126 (Continued )

(Continued ) State

City

MI

Detroit

MN

Minneapolis-St. Paul MS

Jackson

MO

Length of record (yr.)

Jan.

Feb.

Mar.

Apr.

May

June

July

Aug.

29

13

11

13

12

11

Sault Ste. Marie

46

19

15

13

11

Duluth

46

12

10

11

10

49

9

7

10

24

11

9

10

Kansas City

15

7

7

St. Louis

30

8

8

MT

Great Falls

50

9

8

NE

Omaha

51

6

7

Sept.

11

9

9

10

11

12

10

11

12

13

11

11

10

11

12

10

8

10

8

10

11

11

11

11

11

11

11

10

9

9

12

9

10

12

Oct.

Nov.

Dec.

Annual

9

12

14

135

13

13

17

20

165

12

10

11

12

134

10

10

8

8

9

115

10

8

6

8

10

109

7

9

8

8

8

8

107

8

8

8

8

10

9

111

12

7

8

7

6

7

8

101

11

9

9

9

7

5

6

98

NV

Reno

45

6

6

6

4

4

3

2

2

2

3

5

6

51

NH

Concord

46

11

10

11

12

12

11

10

10

9

9

11

11

125

NJ

Atlantic City

44

11

10

11

11

10

9

9

9

8

7

9

10

112

NM

Albuquerque

48

4

4

5

3

4

4

9

9

6

5

3

4

61

NY

Albany

41

12

10

12

12

13

11

10

10

10

9

12

12

134

Buffalo

44

20

17

16

14

12

10

10

11

11

12

16

20

169

New York

118

11

10

11

11

11

10

10

10

8

8

9

10

121

NC

Charlotte

48

10

10

11

9

10

10

11

9

7

7

8

10

111

Raleigh

43

10

10

10

9

10

9

11

10

8

7

8

9

111

ND

Bismarck

48

8

7

8

8

10

12

9

9

7

6

6

8

97

OH

Cincinnati

40

12

11

13

13

11

11

10

9

8

8

11

12

129

Cleveland

46

16

14

15

14

13

11

10

10

10

11

14

16

156 137

Columbus

48

13

12

14

13

13

11

11

9

8

9

11

13

OK

Oklahoma City

48

5

6

7

8

10

9

6

6

7

6

5

5

82

OR

Portland

47

18

16

17

14

12

9

4

5

8

13

18

19

152

PA

Philadelphia

47

11

9

11

11

11

10

9

9

8

8

9

10

117

Pittsburgh

35

16

14

16

14

12

12

11

10

9

11

13

17

154

RI

Providence

34

11

10

12

11

11

11

9

10

8

8

11

12

124

SC

Columbia

40

10

10

11

8

9

9

12

11

8

6

7

9

109

SD

Sioux Falls

42

6

6

9

9

10

11

9

9

8

6

6

6

97

TN TX

Memphis

37

10

9

11

10

9

8

9

8

7

6

9

10

106

Nashville

46

11

11

12

11

11

9

10

9

8

7

10

11

119

Dallas-Fort Worth

34

7

7

7

8

9

6

5

5

7

6

6

6

78

El Paso

48

4

3

2

2

2

4

8

8

5

4

3

4

48

Houston

18

10

8

9

7

9

9

9

10

10

8

9

9

106

Salt Lake City

59

10

9

10

9

8

5

5

6

5

6

8

9

91

VT

Burlington

44

14

12

13

12

14

13

12

12

12

12

14

15

154

VA

Norfolk

39

10

10

11

10

10

9

11

10

8

8

8

9

114

Richmond

50

10

9

11

9

11

9

11

10

8

7

8

9

113

Seattle

43

19

16

17

14

10

9

5

6

9

13

18

20

156

Spokane

40

14

12

11

9

9

8

4

5

6

8

12

15

113

WV

Charleston

40

16

14

15

14

13

11

13

11

9

10

12

14

151

WI

Milwaukee

47

11

10

12

12

12

11

10

9

9

9

10

11

125

WY

Cheyenne

52

6

6

9

10

12

11

11

10

7

6

6

5

99

PR

San Juan

32

16

13

12

13

17

16

19

18

17

17

18

19

195

UT

WA

Source: U.S. National Oceanic and Atmospheric Administration, Comparative Climatic Data.

32

CH A P T E R 2: Describing Data Sets

17. The following give the winning scores of the Masters Golf tournament from 2005 through 2009. Use them in conjunction with data given in Example 2.2 to obtain a relative frequency table of all winning scores from 1990 to 2009. Also, use the data given in Example 2.2 to obtain a relative frequency table of all winning scores from 1970 to 1989. Do winning scores appear to have changed much over the past 20 years? Year Winner

Score

2005 2006 2007 2008 2009

276 281 289 280 276

Tiger Woods Phil Mickelson Zach Johnson Trevor Immelman Angel Cabrera

2.3 GROUPED DATA AND HISTOGRAMS As seen in Sec. 2.2, using a line or a bar graph to plot the frequencies of data values is often an effective way of portraying a data set. However, for some data sets the number of distinct values is too large to utilize this approach. Instead, in such cases, we divide the values into groupings, or class intervals, and then plot the number of data values falling in each class interval. The number of class intervals chosen should be a trade-off between (1) choosing too few classes at a cost of losing too much information about the actual data values in a class and (2) choosing too many classes, which will result in the frequencies of each class being too small for a pattern to be discernible. Although 5 to 10 class intervals are typical, the appropriate number is a subjective choice, and of course you can try different numbers of class intervals to see which of the resulting charts appears to be most revealing about the data. It is common, although not essential, to choose class intervals of equal length. The endpoints of a class interval are called the class boundaries. We will adopt the left-end inclusion convention, which stipulates that a class interval contains its left-end but not its right-end boundary point. Thus, for instance, the class interval 20–30 contains all values that are both greater than or equal to 20 and less than 30. The data in Table 2.5 represent the blood cholesterol levels of 40 ﬁrst-year students at a particular college. As a prelude to determining class size frequencies, it is useful to rearrange the data in increasing order. This gives the 40 values of Table 2.6. Since the data range from a minimum value of 171 to a maximum of 227, the left-end boundary of the ﬁrst class interval must be less than or equal to 171, and the right-end boundary of the ﬁnal class interval must be greater than 227.

2.3 Grouped Data and Histograms

Table 2.5 Blood Cholesterol Levels 213 192 187 216 221

174 200 181 206 212

193 200 193 195 221

196 199 205 191 204

220 178 196 171 204

183 183 211 194 191

194 188 202 184 183

200 193 213 191 227

Table 2.6 Blood Cholesterol Levels in Increasing Order 171, 174, 178, 181, 183, 183, 183, 184, 187, 188, 191, 191, 191, 192, 193, 193, 193, 194, 194, 195, 196, 196, 199, 200, 200, 200, 202, 204, 204, 205, 206, 211, 212, 213, 213, 216, 220, 221, 221, 227

Table 2.7 Frequency Table of Blood Cholesterol Levels Class intervals

Frequency

170–180

3

180–190

7

190–200

13

200–210

8

210–220

5

220–230

4

Relative frequency 3 = 0.075 40 7 = 0.175 40 13 = 0.325 40 8 = 0.20 40 5 = 0.125 40 4 = 0.10 40

One choice would be to have the ﬁrst class interval be 170 to 180. This will result in six class intervals. A frequency table giving the frequency (as well as the relative frequency) of data values falling in each class interval is seen in Table 2.7. Note: Because of the left-end inclusion convention, the values of 200 were placed in the class interval of 200 to 210, not in the interval of 190 to 200. A bar graph plot of the data, with the bars placed adjacent to each other, is called a histogram. The vertical axis of a histogram can represent either the class frequency or the relative class frequency. In the former case, the histogram is called a

33

34

CH A P T E R 2: Describing Data Sets

frequency histogram and in the latter a relative frequency histogram. Figure 2.7 presents a frequency histogram of the data of Table 2.7. It is important to recognize that a class frequency table or a histogram based on that table does not contain all the information in the original data set. These two representations note only the number of data values in each class and not the actual data values themselves. Thus, whereas such tables and charts are useful for illustrating data, the original raw data set should always be saved. To Construct a Histogram from a Data Set 1. Arrange the data in increasing order. 2. Choose class intervals so that all data points are covered. 3. Construct a frequency table. 4. Draw adjacent bars having heights determined by the frequencies in step 3. The importance of a histogram is that it enables us to organize and present data graphically so as to draw attention to certain important features of the data. For instance, a histogram can often indicate 1. 2. 3. 4. 5.

How symmetric the data are How spread out the data are Whether there are intervals having high levels of data concentration Whether there are gaps in the data Whether some data values are far apart from others

FIGURE 2.7 Frequency histogram for the data of Table 2.7.

2.3 Grouped Data and Histograms

FIGURE 2.8 Characteristics of data detected by histograms. (a) symmetry, (b) degree of spread and where values are concentrated, and (c) gaps in data and data far from others.

For instance, the histogram presented in Fig. 2.7 indicates that the frequencies of the successive classes ﬁrst increase and then decrease, reaching a maximum in the class having limits of 190 to 200. The histograms of Fig. 2.8 give valuable information about the data sets they represent. The data set whose histogram is on the left side of Fig. 2.8(a) is symmetric, whereas the one on the right side is not. The data set represented on the left side of Fig. 2.8(b) is fairly evenly spread out, whereas the one for the right side is more concentrated. The data set represented by the left side of Fig. 2.8(c) has a gap, whereas the one represented on the right side has certain values far apart from the rest. ■

Example 2.3 Table 2.8 gives the birth rates (per 1000 population) in each of the 50 states of the United States. Plot these data in a histogram.

35

36

CH A P T E R 2: Describing Data Sets

Table 2.8 Birth Rates per 1000 Population State

Rate

State

Rate

State

Rate

Alabama

14.2

Louisiana

15.7

Ohio

14.9

Alaska

21.9

Maine

13.8

Oklahoma

14.4

Arizona

19.0

Maryland

14.4

Oregon

15.5

Arkansas

14.5

Massachusetts

16.3

Pennsylvania

14.1

California

19.2

Michigan

15.4

Rhode Island

15.3

Colorado

15.9

Minnesota

15.3

South Carolina

15.7

Connecticut

14.7

Mississippi

16.1

South Dakota

15.4

Delaware

17.1

Missouri

15.5

Tennessee

15.5

Florida

15.2

Montana

14.1

Texas

17.7

Georgia

17.1

Nebraska

15.1

Utah

21.2

Hawaii

17.6

Nevada

16.5

Vermont

14.0

Idaho

15.2

New Hampshire

16.2

Virginia

15.3

Illinois

16.0

New Jersey

15.1

Washington

15.4

Indiana

14.8

New Mexico

17.9

West Virginia

12.4

Iowa

13.1

New York

16.2

Wisconsin

14.8

Wyoming

13.7

Kansas

14.2

North Carolina

15.6

Kentucky

14.1

North Dakota

16.5

Source: Department of Health and Human Services.

Solution Since the data range from a low value of 12.4 to a high of 21.9, let us use class intervals of length 1.5, starting at the value 12. With these class intervals, we obtain the following frequency table. Class intervals Frequency Class intervals Frequency 12.0–13.5

2

18.0–19.5

2

13.5–15.0

15

19.5–21.0

0

15.0–16.5

22

21.0–22.5

2

16.5–18.0

7

A histogram plot of these data is presented in Fig. 2.9. A histogram is, in essence, a bar chart that graphs the frequencies or relative frequencies of data falling into different class intervals. These class frequencies can also be represented graphically by a frequency (or relative frequency)

2.3 Grouped Data and Histograms

FIGURE 2.9 A histogram for birth rates in the 50 states.

polygon. Each class interval is represented by a value, usually taken to be the midpoint of that interval. A plot is made of these values versus the frequencies of the class intervals they represent. These plotted points are then connected by straight lines to yield the frequency polygon. Such graphs are particularly useful for comparing data sets, since the different frequency polygons can be plotted on the same chart. ■ ■

Example 2.4 The data of Table 2.9 represent class frequencies for the systolic blood pressure of two groups of male industrial workers: those aged 30 to 40 and those aged 50 to 60. It is difﬁcult to directly compare the blood pressures for the two age groups since the total number of workers in each group is different. To remove this difﬁculty, we can compute and graph the relative frequencies of each of the classes. That is, we divide all the frequencies relating to workers aged 30 to 39 by 2540 (the number of such workers) and all the frequencies relating to workers aged 50 to 59 by 731. This results in Table 2.10. Figure 2.10 graphs the relative frequency polygons for both age groups. Having both frequency polygons on the same graph makes it easy to compare the two data sets. For instance, it appears that the blood pressures of the older group are more spread out among larger values than are those of the younger group. ■

37

38

CH A P T E R 2: Describing Data Sets

Table 2.9 Class Frequencies of Systolic Blood Pressure of Two Groups of Male Workers Number of workers Blood pressure Less than 90 90–100 100–110 110–120 120–130 130–140 140–150 150–160 160–170 170–180 180–190 190–200 200–210 210–220 220–230 230–240 Total

Aged 30–40

Aged 50–60

3 17 118 460 768 675 312 120 45 18 3 1

1 2 23 57 122 149 167 73 62 35 20 9 3 5 2 1

2540

731

Table 2.10 Relative Class Frequencies of Blood Pressures Percentage of workers Blood pressure

Aged 30–40

Aged 50–60

Less than 90 90–100 100–110 110–120 120–130 130–140 140–150 150–160 160–170 170–180 180–190 190–200 200–210 210–220 220–230 230–240

0.12 0.67 4.65 18.11 30.24 26.57 12.28 4.72 1.77 0.71 0.12 0.04

0.14 0.27 3.15 7.80 16.69 20.38 22.84 9.99 8.48 4.79 2.74 1.23 0.41 0.68 0.27 0.14

Total

100.00

100.00

FIGURE 2.10 Relative frequency polygons for the data of Table 2.10.

2.3 Grouped Data and Histograms

PROBLEMS 1. The following data set represents the scores on intelligence quotient (IQ) examinations of 40 sixth-grade students at a particular school: 114, 122, 103, 118, 99, 105, 134, 125, 117, 106, 109, 104, 111, 127, 133, 111, 117, 103, 120, 98, 100, 130, 141, 119, 128, 106, 109, 115, 113, 121, 100, 130, 125, 117, 119, 113, 104, 108, 110, 102 (a) (b) (c) (d)

Present this data set in a frequency histogram. Which class interval contains the greatest number of data values? Is there a roughly equal number of data in each class interval? Does the histogram appear to be approximately symmetric? If so, about which interval is it approximately symmetric? 2. The following data represent the daily high temperature (in degrees Celsius) on July 4 in San Francisco over a sequence of 30 years: 22.8, 26.2, 31.7, 31.1, 26.9, 28.0, 29.4, 28.8, 26.7, 27.4, 28.2, 30.3, 29.5, 28.9, 27.5, 28.3, 24.1, 25.3, 28.5, 27.7, 24.4, 29.2, 30.3, 33.7, 27.5, 29.3, 30.2, 28.5, 32.2, 33.7 (a) Present this data set in a frequency histogram. (b) What would you say is a “typical” July 4 temperature in San Francisco? (c) What other conclusions can be drawn from the histogram? 3. The following data (in thousands of dollars) represent the net annual income for a sample of taxpayers: 47, 55, 18, 24, 27, 41, 50, 38, 33, 29, 15, 77, 64, 22, 19, 35, 39, 41, 67, 55, 121, 77, 80, 34, 41, 48, 60, 30, 22, 28, 84, 55, 26, 105, 62, 30, 17, 23, 31, 28, 56, 64, 88, 104, 115, 39, 25, 18, 21, 30, 57, 40, 38, 29, 19, 46, 40, 49, 72, 70, 37, 39, 18, 22, 29, 52, 94, 86, 23, 36 (a) Graph this data set in a frequency histogram having 5 class intervals. (b) Graph this data set in a frequency histogram having 10 class intervals. (c) Which histogram do you think is more informative? Why? 4. A set of 200 data points was broken up into 8 classes each of size (in the units of the data) 3, and the frequency of values in each class was determined. A frequency table was then constructed. However,

39

40

CH A P T E R 2: Describing Data Sets

some of the entries of this table were lost. Suppose that the part of the frequency table that remains is as follows: Class interval Frequency Relative frequency 0.05 14 18 38

15–18

0.10 42 11

Fill in the missing numbers and draw a relative frequency histogram. 5. The following is the ozone concentration (measured in parts per 100 million) of air in the downtown Los Angeles area during 25 consecutive summer days in 2004: 6.2, 9.1, 2.4, 3.6, 1.9, 1.7, 4.5, 4.2, 3.3, 5.1, 6.0, 1.8, 2.3, 4.9, 3.7, 3.8, 5.5, 6.4, 8.6, 9.3, 7.7, 5.4, 7.2, 4.9, 6.2 (a) Construct a frequency histogram for this data set having 3 to 5 as a class interval. (b) Construct a frequency histogram for this data set having 2 to 3 as a class interval. (c) Which frequency histogram do you ﬁnd more informative? 6. The following is the 2002 meat production, in thousands of metric tons, for 11 different countries. Country Argentina Australia Brazil China France Italy

Production Country 2,748 2,034 7,150 5,616 1,666 1,161

Japan Mexico Spain United Kingdom United States

Production 520 1, 450 592 1, 390 12, 424

(a) Represent the given data in a frequency histogram. (b) A data value that is far removed from the others is called an outlier. Is there an outlier in the given data? 7. Consider the blood cholesterol levels of the ﬁrst 100 students in the data set presented in App. A. Divide these students by gender groupings, and construct a class relative frequency table for each. Plot, on the same chart, separate class relative frequency polygons for the

2.3 Grouped Data and Histograms

female and male students. Can any conclusions be drawn about the relationship between gender and cholesterol level? 8. Use the following table to construct a frequency histogram of the 2008 state sales tax rates of the 50 states and the District of Columbia. State Sales Tax Rates, January 1, 2008 State

Tax rates

Alabama

4

Alaska

none

Arizona

5.6

State

Tax rates

State

Tax rates

Louisiana

4

Ohio

5.5

Maine

5

Oklahoma

4.5

Maryland

6

Oregon

Massachusetts

5

Pennsylvania

6

Rhode Island

7

none

Arkansas

6

California

7.25

Michigan

6

Colorado

2.9

Minnesota

6.5

South Carolina

6

6

Mississippi

7

South Dakota

4

Connecticut Delaware

none

Missouri

4.225

Tennessee

Florida

6

Montana

none

Texas

6.25 4.65

7

Georgia

4

Nebraska

5.5

Utah

Hawaii

4

Nevada

6.5

Vermont

6

Idaho

6

New Hampshire

none

Virginia

5

Illinois

6.25

New Jersey

7

Washington

6.5

Indiana

6

New Mexico

5

West Virginia

6

Iowa

5

New York

4

Wisconsin

5

Wyoming

Kansas Kentucky

5.3

North Carolina

4.25

6

North Dakota

5

4

Dist. of Columbia

5.75

The following table provides data concerning accidental death rates in the United States over a variety of years. Use it to answer Problems 9 through 12. Death Rates per 100,000 Population for the Principal Types of Accidental Deaths in the United States, 1970–2002

Year

Motor vehicle

Falls

Poisoning

Drowning

Fires, ﬂames, smoke

1970 1980 1985 1990 1991 1992 1993

26.8 23.4 19.3 18.8 17.3 16.1 16.3

8.3 5.9 5.0 4.9 5.0 5.0 5.1

2.6 1.9 2.2 2.3 2.6 2.7 3.4

3.9 3.2 2.2 1.9 1.8 1.4 1.5

3.3 2.6 2.1 1.7 1.6 1.6 1.5

Ingestion of food, object

Firearms

1.4 1.4 1.5 1.3 1.3 1.2 1.2

1.2 0.9 0.7 0.6 0.6 0.6 0.6 (Continued)

41

42

CH A P T E R 2: Describing Data Sets

(Continued)

Year

Motor vehicle

Falls

Poisoning

Drowning

Fires, ﬂames, smoke

1994 1995 1996 1997 1998 1999 2000 2001 2002

16.3 16.5 16.5 16.2 16.1 15.5 15.7 15.7 15.7

5.2 5.3 5.6 5.8 6.0 4.8 4.8 5.1 5.2

3.5 3.4 3.5 3.8 4.0 4.5 4.6 5.0 5.6

1.5 1.7 1.5 1.5 1.6 1.3 1.3 1.2 1.1

1.5 1.4 1.4 1.3 1.2 1.2 1.2 1.2 1.0

Ingestion of food, object

Firearms

1.2 1.2 1.2 1.2 1.3 1.4 1.6 1.4 1.5

0.5 0.5 0.4 0.4 0.3 0.3 0.3 0.3 0.3

Source: National Safety Council.

9. Construct a relative frequency histogram of yearly death rates due to motor vehicles. 10. Construct a relative frequency histogram of yearly death rates due to falls. 11. Construct a relative frequency histogram of total yearly death rates due to all listed causes. 12. Would you say that the accidental death rates are remaining relatively steady? 13. Using the table described prior to Prob. 12 in Sec. 2.2, construct a histogram for the average yearly number of rainy days for the cities listed. 14. Consider the following table.

Age of driver, years 15–20 20–25 25–30 30–35 35–40 40–45 45–50 50–55 55–60 60–65 65–70 70–75 Over 75

Percentage of all drivers 9 13 13 11 9 8 8 7 6 6 4 3 3

Percentage of all drivers in fatal accidents 18 21 14 11 7 6 5 5 4 3 2 2 2

2.3 Grouped Data and Histograms

By the left-end convention, 13 percent of all drivers are at least 25 but less than 30 years old, and 11 percent of drivers killed in car accidents are at least 30 but less than 35 years old. (a) Draw a relative frequency histogram for the age breakdown of drivers. (b) Draw a relative frequency histogram for the age breakdown of those drivers who are killed in car accidents. (c) Which age group accounts for the largest number of fatal accidents? (d) Which age group should be charged the highest insurance premiums? Explain your reasoning. 15. A cumulative relative frequency table gives, for an increasing sequence of values, the percentage of data values that are less than that value. It can be constructed from a relative frequency table by simply adding the relative frequencies in a cumulative fashion. The following table is the beginning of such a table for the two data sets shown in Table 2.9. It says, for instance, that 5.44 percent of men aged 30 to 40 years have blood pressures below 110, as opposed to only 3.56 percent of those aged 50 to 60 years. A Cumulative Relative Frequency Table for the Data Sets of Table 2.9 Percentage of workers Blood pressure less than Aged 30–40 Aged 50–60 90 100 110 120 130 · · · 240

0.12 0.79 5.44

100

0.14 0.41 3.56

100

(a) Explain why the cumulative relative frequency for the last class must be 100. (b) Complete the table. (c) What does the table tell you about the two data sets? (That is, which one tends to have smaller values?) (d) Graph, on the same chart, cumulative relative frequency polygons for the given data. Such graphs are called ogives (pronounced “OH jives”).

43

44

CH A P T E R 2: Describing Data Sets

2.4 STEM-AND-LEAF PLOTS A very efﬁcient way of displaying a small-to-moderate size data set is to utilize a stem-and-leaf plot. Such a plot is obtained by dividing each data value into two parts—its stem and its leaf. For instance, if the data are all two-digit numbers, then we could let the stem of a data value be the tens digit and the leaf be the ones digit. That is, the value 84 is expressed as Stem Leaf 4 8 and the two data values 84 and 87 are expressed as Stem Leaf 4, 7 8

■

Example 2.5 Table 2.11 presents the per capita personal income for each of the 50 states and the District of Columbia. The data are for 2002.

Table 2.11 Per Capita Personal Income (Dollars per Person), 2002 State name

State name

State name

United States

30,941

Kentucky

25,579

Ohio

29,405

Alabama

25,128

Louisiana

25,446

Oklahoma

25,575

Alaska

32,151

Maine

27,744

Oregon

28,731

Arizona

26,183

Maryland

36,298

Pennsylvania

31,727

Arkansas

23,512

Massachusetts

39,244

Rhode Island

31,319

California

32,996

Michigan

30,296

South Carolina

25,400

Colorado

33,276

Minnesota

34,071

South Dakota

26,894

Connecticut

42,706

Mississippi

22,372

Tennessee

27,671

Delaware

32,779

Missouri

28,936

Texas

28,551

District of Columbia

42,120

Montana

25,020

Utah

24,306

Florida

29,596

Nebraska

29,771

Vermont

29,567

Georgia

28,821

Nevada

30,180

Virginia

32,922

Hawaii

30,001

New Hampshire

34,334

Washington

32,677

Idaho

25,057

New Jersey

39,453

West Virginia

23,688

Illinois

33,404

New Mexico

23,941

Wisconsin

29,923

Indiana

28,240

New York

36,043

Wyoming

30,578

Iowa

28,280

North Carolina

27,711

Kansas

29,141

North Dakota

26,982

2.4 Stem-and-Leaf Plots

The data presented in Table 2.11 are represented in the following stem-and-leaf plot. Note that the values of the leaves are put in the plot in increasing order. 372 512, 688, 941 706 020, 057, 128, 400, 446, 575, 579 183, 894, 982 671, 711, 744 240, 280, 551, 731, 821, 936 141, 405, 567, 596, 771, 923 001, 180, 296, 578 319, 727 151, 677, 779, 922, 996 276, 404 071, 334 043, 298 244, 453 120, 706

22 23 24 25 26 27 28 29 30 31 32 33 34 36 39 42

The choice of stems should always be made so that the resultant stem-and-leaf plot is informative about the data. For instance, consider Example 2.6. ■ ■

Example 2.6 The following data represent the proportion of public elementary school students that are classiﬁed as minority in each of 18 cities. 55.2, 47.8, 44.6, 64.2, 61.4, 36.6, 28.2, 57.4, 41.3, 44.6, 55.2, 39.6, 40.9, 52.2, 63.3, 34.5, 30.8, 45.3 If we let the stem denote the tens digit and the leaf represent the remainder of the value, then the stem-and-leaf plot for the given data is as follows: 2 3 4 5 6

8.2 0.8, 4.5, 6.6, 9.6 0.9, 1.3, 4.6, 4.6, 5.3, 7.8 2.2, 5.2, 5.2, 7.4 1.4, 3.3, 4.2

We could have let the stem denote the integer part and the leaf the decimal part of the value, so that the value 28.2 would be represented as 28 .2

45

46

CH A P T E R 2: Describing Data Sets

However, this would have resulted in too many stems (with too few leaves each) to clearly illustrate the data set. ■ ■

Example 2.7 The following stem-and-leaf plot represents the weights of 80 attendees at a sporting convention. The stem represents the tens digit, and the leaves are the ones digit. 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

2, 3, 3, 4, 7 0, 1, 2, 2, 3, 6, 9 1, 2, 4, 4, 6, 6, 6, 7, 9 1, 2, 2, 5, 5, 6, 6, 8, 9 0, 4, 6, 7, 7, 9, 9 1, 1, 5, 6, 6, 6, 7 0, 1, 1, 1, 2, 4, 5, 6, 8, 8 1, 1, 3, 5, 6, 6, 6 1, 2, 2, 5, 5, 6, 6, 9 0, 0, 1, 2, 4, 5 9, 9 7 1 9

(5) (7) (9) (9) (7) (7) (10) (7) (8) (6) (2) (1) (1) (0) (1)

The numbers in parentheses on the right represent the number of values in each stem class. These summary numbers are often useful. They tell us, for instance, that there are 10 values having stem 16; that is, 10 individuals have weights between 160 and 169. Note that a stem without any leaves (such as stem value 23) indicates that there are no occurrences in that class. It is clear from this plot that almost all the data values are between 100 and 200, and the spread is fairly uniform throughout this region, with the exception of fewer values in the intervals between 100 and 110 and between 190 and 200. ■ Stem-and-leaf plots are quite useful in showing all the data values in a clear representation that can be the ﬁrst step in describing, summarizing, and learning from the data. It is most helpful in moderate-size data sets. (If the size of the data set were very large, then, from a practical point of view, the values of all the leaves might be too overwhelming and a stem-and-leaf plot might not be any more informative than a histogram.) Physically this plot looks like a histogram turned on its side, with the additional plus that it presents the original within-group data values. These within-group values can be quite valuable to help you discover patterns in the data, such as that all the data values are multiples of some common value, or ﬁnd out which values occur most frequently within a stem group.

2.4 Stem-and-Leaf Plots

Sometimes a stem-and-leaf plot appears to have too many leaves per stem line and as a result looks cluttered. One possible solution is to double the number of stems by having two stem lines for each stem value. On the top stem line in the pair we could include all leaves having values 0 through 4, and on the bottom stem line all leaves having values 5 through 9. For instance, suppose one line of a stem-and-leaf plot is as follows: 6 0, 0, 1, 2, 2, 3, 4, 4, 4, 4, 5, 5, 6, 6, 7, 7, 7, 7, 8, 9, 9 This could be broken into two lines: 6 0, 0, 1, 2, 2, 3, 4, 4, 4, 4 6 5, 5, 6, 6, 7, 7, 7, 7, 8, 9, 9

PROBLEMS 1. For the following data, draw stem-and-leaf plots having (a) 4 stems and (b) 8 stems. 124, 129, 118, 135, 114, 139, 127, 141, 111, 144, 133, 127, 122, 119, 132, 137, 146, 122, 119, 115, 125, 132, 118, 126, 134, 147, 122, 119, 116, 125, 128, 130, 127, 135, 122, 141 2. The following table gives the maximal marginal 2008 tax rates of a variety of states. Represent the data in a stem and leaf plot. State Individual Income Taxes (Tax rates for tax year 2008 – as of January 1, 2008) State Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Hawaii

Maximal rate 5 0 4.54 7.0 9.3 4.63 5.0 5.95 0 6.0 8.25

State Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts

Source: Statistical Abstract of the United States.

Maximal rate 7.8 3.0 3.4 8.98 6.45 6.0 6.0 8.5 5.5 5.3

47

48

CH A P T E R 2: Describing Data Sets

3. The following are the ages, to the nearest year, of 43 patients admitted to the emergency ward of a certain adult hospital: 23, 18, 31, 79, 44, 51, 24, 19, 17, 25, 27, 19, 44, 61, 22, 18, 14, 17, 29, 31, 22, 17, 15, 40, 55, 16, 17, 19, 20, 32, 20, 45, 53, 27, 16, 19, 22, 20, 18, 30, 20, 33, 21 Draw a stem-and-leaf plot for this data set. Use this plot to determine the 5-year interval of ages that contains the largest number of data points. 4. A psychologist recorded the following 48 reaction times (in seconds) to a certain stimulus. 1.1, 2.1, 0.4, 3.3, 1.5, 1.3, 3.2, 2.0, 1.7, 0.6, 0.9, 1.6, 2.2, 2.6, 1.8, 0.9, 2.5, 3.0, 0.7, 1.3, 1.8, 2.9, 2.6, 1.8, 3.1, 2.6, 1.5, 1.2, 2.5, 2.8, 0.7, 2.3, 0.6, 1.8, 1.1, 2.9, 3.2, 2.8, 1.2, 2.4, 0.5, 0.7, 2.4, 1.6, 1.3, 2.8, 2.1, 1.5 (a) (b) (c) (d)

Construct a stem-and-leaf plot for these data. Construct a second stem-and-leaf plot, using additional stems. Which one seems more informative? Suppose a newspaper article stated, “The typical reaction time was ______ seconds.” Fill in your guess as to the missing word. 5. The following data represent New York City’s daily revenue from parking meters (in units of $5000) during 30 days in 2002. 108, 77, 58, 88, 65, 52, 104, 75, 80, 83, 74, 68, 94, 97, 83, 71, 78, 83, 90, 79, 84, 81, 68, 57, 59, 32, 75, 93, 100, 88 (a) Represent this data set in a stem-and-leaf plot. (b) Do any of the data values seem “suspicious”? Why? 6. The volatility of a stock is an important property in the theory of stock options pricing. It is an indication of how much change there tends to be in the day-to-day price of the stock. A volatility of 0 means that the price of the stock always remains the same. The higher the volatility, the more the stock’s price tends to change. The following is a list of the volatility of 32 companies whose stock is traded on the American Stock Exchange: 0.26, 0.31, 0.45, 0.30, 0.26, 0.17, 0.33, 0.32, 0.37, 0.38, 0.35, 0.28, 0.37, 0.35, 0.29, 0.20, 0.33, 0.19, 0.31, 0.26, 0.24, 0.50, 0.22, 0.33, 0.51, 0.44, 0.63, 0.30, 0.28, 0.48, 0.42, 0.37

2.4 Stem-and-Leaf Plots

(a) Represent these data in a stem-and-leaf plot. (b) What is the largest data value? (c) What is the smallest data value? (d) What is a “typical” data value? 7. The following table gives the scores of the ﬁrst 25 Super Bowl games in professional football. Use it to construct a stem-and-leaf plot of (a) The winning scores (b) The losing scores (c) The amounts by which the winning teams outscored the losing teams Super Bowls I-XXV Game

Date

Winner

Loser

XXV

Jan. 27, 1991

New York (NFC) 20

Buffalo (AFC) 19

XXIV

Jan. 28, 1990

San Francisco (NFC) 55

Denver (AFC) 10

XXIII

Jan. 22, 1989

San Francisco (NFC) 20

Cincinnati (AFC) 16

XXII

Jan. 31, 1988

Washington (NFC) 42

Denver (AFC) 10

XXI

Jan. 25, 1987

New York (NFC) 39

Denver (AFC) 20

XX

Jan. 26, 1986

Chicago (NFC) 46

New England (AFC) 10

XIX

Jan. 20, 1985

San Francisco (NFC) 38

Miami (AFC) 16

XVIII

Jan. 22, 1984

Los Angeles Raiders (AFC) 38

Washington (NFC) 9

XVII

Jan. 30, 1983

Washington (NFC) 27

Miami (AFC) 17

XVI

Jan. 24, 1982

San Francisco (NFC) 26

Cincinnati (AFC) 21

XV

Jan. 25, 1981

Oakland (AFC) 27

Philadelphia (NFC) 10

XIV

Jan. 20, 1980

Pittsburgh (AFC) 31

Los Angeles (NFC) 19

XIII

Jan. 21, 1979

Pittsburgh (AFC) 35

Dallas (NFC) 31

XII

Jan. 15, 1978

Dallas (NFC) 27

Denver (AFC) 10

XI

Jan. 9, 1977

Oakland (AFC) 32

Minnesota (NFC) 14

X

Jan. 18, 1976

Pittsburgh (AFC) 21

Dallas (NFC) 17

IX

Jan. 12, 1975

Pittsburgh (AFC) 16

Minnesota (NFC) 6

VIII

Jan. 13, 1974

Miami (AFC) 24

Minnesota (NFC) 7

VII

Jan. 14, 1973

Miami (AFC) 14

Washington (NFC) 7

VI

Jan. 16, 1972

Dallas (NFC) 24

Miami (AFC) 3

V

Jan. 17, 1971

Baltimore (AFC) 16

Dallas (NFC) 13

IV

Jan. 11, 1970

Kansas City (AFL) 23

Minnesota (NFL) 7

III

Jan. 12, 1969

New York (AFL) 16

Baltimore (NFL) 7

II

Jan. 14, 1968

Green Bay (NFL) 33

Oakland (AFL) 14

I

Jan. 15, 1967

Green Bay (NFL) 35

Kansas City (AFL) 10

49

50

CH A P T E R 2: Describing Data Sets

8. Consider the following stem-and-leaf plot and histogram concerning the same set of data. 2 3 4 5 6 7

1,1,4,7 0, 0, 3, 3, 6, 9, 9, 9 2, 2, 5, 8, 8, 8 1, 1, 7, 7 3, 3, 3, 6 2, 2, 5, 5, 5, 8

2–3 3–4 4–5 5–6 6–7 7–8

x, x, x, x x, x, x, x, x, x, x, x x, x, x, x, x, x x, x, x, x x, x, x, x x, x, x, x, x, x

What can you conclude from the stem-and-leaf plot that would not have been apparent from the histogram? 9. Use the data represented in the stem-and-leaf plot in Prob. 8 to answer the following questions. (a) How many data values are in the 40s? (b) What percentage of values is greater than 50? (c) What percentage of values has the ones digit equal to 1? 10. The following table gives the different 2002 incomes and Social Security tax rates for a variety of countries. (a) Represent the percentages paid in income tax in a histogram. (b) Represent the percentages paid in Social Security tax in a stemand-leaf plot.

Tax Burden in Selected Countries*

Country Denmark Belgium Germany Finland Poland Sweden Turkey Netherlands Norway Austria Hungary Italy France Canada Australia ∗

Income tax (%)

Social Security (%)

Total payment† (%)

33 28 21 26 6 23 15 7 21 11 17 19 13 19 24

11 14 21 6 25 7 15 22 8 18 13 9 13 7 0

43 41 41 32 31 30 30 29 29 29 29 28 27 26 24

Country Czech Republic United States United Kingdom Iceland Luxembourg Switzerland New Zealand Slovak Republic Spain Greece Portugal Ireland Japan Korea Mexico

Income tax (%)

Social Security (%)

11 17 16 22 8 10 20 7 13 1 6 11 6 2 2

13 8 8 0 14 12 0 13 6 16 11 5 10 7 2

Total payment† (%)

Does not include taxes not listed, such as sales tax or VAT. Rates shown apply to a single person with average earnings. Totals may not add due to rounding. Source: Organization for Economic Cooperation and Development, 2002. †

24 24 23 22 22 22 20 19 19 17 17 16 16 9 4

2.5 Sets of Paired Data

11. A useful way of comparing two data sets is to put their stem-and-leaf plots side by side. The following represents the scores of students in two different schools on a standard examination. In both schools 24 students took the examination. School A Leaves 0 8, 5 9, 7, 4, 2, 0 9, 8, 8, 7, 7, 6, 5, 3 8, 8, 6, 6, 5, 5, 3, 0

(a) (b) (c) (d)

Stem

School B Leaves

5 6 7 8 9 10

3, 5, 7 2, 5, 8, 9, 9 3, 6, 7, 8, 8, 9 0, 2, 3, 5, 6, 6 0, 1, 5 0

Which school had the “high scorer”? Which school had the “low scorer”? Which school did better on the examination? Combine the two schools, and draw a stem-and-leaf plot for all 48 values.

2.5 SETS OF PAIRED DATA Sometimes a data set consists of pairs of values that have some relationship to each other. Each member of the data set is thought of as having an x value and a y value. We often express the ith pair by the notation (xi , yi ), i = 1, . . . , n. For instance, in the data set presented in Table 2.12, xi represents the score on an intelligence quotient (IQ) test, and yi represents the annual salary (to the nearest $1000) of the ith chosen worker in a sample of 30 workers from a particular company. In this section, we show how to effectively display data sets of paired values. One approach to representing such a data set is to ﬁrst consider each part of the paired data separately and then plot the relevant histograms or stem-and-leaf plots for each. For instance, Figs. 2.11 and 2.12 are stem-and-leaf plots of, respectively, the IQ test scores and the annual salaries for the data presented in Table 2.12. However, although Figs. 2.11 and 2.12 tell us a great deal about the individual IQ scores and worker salaries, they tell us nothing about the relationship between these two variables. Thus, for instance, by themselves they would not be useful in helping us learn whether higher IQ scores tend to go along with higher income at this company. To learn about how the data relate to such questions, it is necessary to consider the paired values of each data point simultaneously. A useful way of portraying a data set of paired values is to plot the data on a two-dimensional rectangular plot with the x axis representing the x value of the data and the y axis representing the y value. Such a plot is called a scatter diagram. Figure 2.13 presents a scatter diagram for the data of Table 2.12.

51

52

CH A P T E R 2: Describing Data Sets

Table 2.12 Salaries versus IQ Worker i 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

12 11 10 9 8 7

IQ score xi

Annual salary yi (in units of $1000)

110 107 83 87 117 104 110 118 116 94 93 101 93 76 91

68 30 13 24 40 22 25 62 45 70 15 22 18 20 14

4 (1) 0,0,2,3,3,5,6,6,7,8 (10) 1,4,4,7 (4) 0,1,1,3,3,4,4,5 (8) 0,3,3,4,7 (5) 6,9 (2)

FIGURE 2.11 Stem-and-leaf plot for IQ scores.

Worker i 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

7 6 5 4 3 2 1

IQ score x i

Annual salary yi (in units of $1000)

84 83 112 80 91 113 124 79 116 113 94 95 104 115 90

19 16 52 11 13 29 71 19 43 44 17 15 30 63 16

0,1 (2) 2,3,8 (3) 2 (1) 0,3,4,5 (4) 0,0 (2) 0,2,2,4,5,9 (6) 1,3,3,4,5,5,6,6,7,8,9,9 (12)

FIGURE 2.12 Stem-and-leaf plot for annual salaries (in $1000).

It is clear from Fig. 2.13 that higher incomes appear to go along with higher scores on the IQ test. That is, while not every worker with a high IQ score receives a larger salary than another worker with a lower score (compare worker 5 with worker 29), it appears to be generally true. The scatter diagram of Fig. 2.13 also appears to have some predictive uses. For instance, suppose we wanted to predict the salary of a worker, similar to the ones just considered, whose IQ test score is 120. One way to do this is to “ﬁt by eye” a line to the data set, as is done in Fig. 2.14. Since the y value on the line

2.5 Sets of Paired Data

FIGURE 2.13 Scatter diagram of IQ versus income data.

FIGURE 2.14 Scatter diagram for IQ versus income: ﬁtting a straight line by eye.

53

54

CH A P T E R 2: Describing Data Sets

corresponding to the x value of 120 is about 45, this seems like a reasonable prediction for the annual salary of a worker whose IQ is 120. In addition to displaying joint patterns of two variables and guiding predictions, a scatter diagram is useful in detecting outliers, which are data points that do not appear to follow the pattern of the other data points. (For example, the point (94, 70) in Fig. 2.13 does not appear to follow the general trend.) Having noted the outliers, we can then decide whether the data pair is meaningful or is caused by an error in data collection.

PROBLEMS 1. In an attempt to determine the relationship between the daily midday temperature (measured in degrees Celsius) and the number of defective parts produced during that day, a company recorded the following data over 22 workdays. Number of Temperature defective parts

Number of Temperature defective parts

24.2

25

24.8

23

22.7

31

20.6

20

30.5

36

25.1

25

28.6

33

21.4

25

25.5

19

23.7

23

32.0

24

23.9

27

28.6

27

25.2

30

26.5

25

27.4

33

25.3

16

28.3

32

26.0

14

28.8

35

24.4

22

26.6

24

(a) Draw a scatter diagram. (b) What can you conclude from the scatter diagram? (c) If tomorrow’s midday temperature reading were 24.0, what would your best guess be as to the number of defective parts produced? 2. The following table gives, for each state, the percentage of its population not covered by health insurance, in the years 1990, 2000, and 2002. (a) Draw a scatter diagram relating the 1990 and 2000 rates. (b) Draw a scatter diagram relating the 2000 and 2002 rates.

Health Insurance Coverage* by State, 1990, 2000, 2002 2002

2000

2002

1990

Not % not Not % not Not % not covered† covered covered† covered covered† covered AL AK AZ AR CA CO CT DE DC FL GA HI ID IL IN IA KS KY LA ME MD MA MI MN MS MO ∗

564 119 916 440 6,398 720 356 79 74 2,843 1,354 123 233 1,767 797 277 280 548 820 144 730 644 1,158 397 465 646

12.7 18.7 16.8 16.3 18.2 16.1 10.5 9.9 13.0 17.3 16.1 10.0 17.9 14.1 13.1 9.5 10.4 13.6 18.4 11.3 13.4 9.9 11.7 7.9 16.7 11.6

582 117 869 379 6,299 620 330 72 78 2,829 1,166 113 199 1,704 674 253 289 545 789 138 547 549 901 399 380 524

13.3 18.7 16.7 14.3 18.5 14.3 9.8 9.3 14.0 17.7 14.3 9.4 15.4 13.9 11.2 8.8 10.9 13.6 18.1 10.9 10.4 8.7 9.2 8.1 13.6 9.5

710 77 547 421 5,683 495 226 96 109 2,376 971 81 159 1,272 587 225 272 480 797 139 601 530 865 389 531 665

17.4 15.4 15.5 17.4 19.1 14.7 6.9 13.9 19.2 18.0 15.3 7.3 15.2 10.9 10.7 8.1 10.8 13.2 19.7 11.2 12.7 9.1 9.4 8.9 19.9 12.7

1990

Not % not Not % not Not % not covered† covered covered† covered covered† covered MT NE NV NH NJ NM NY NC ND OH OK OR PA RI SC SD TN TX UT VT VA WA WV WI WY U.S.

139 174 418 125 1,197 388 3,042 1,368 69 1,344 601 511 1,380 104 500 85 614 5,556 310 66 962 850 255 538 86 43,574

For population, all ages, including those 65 or over, an age group largely covered by Medicare. In thousands. Source: Bureau of the Census. U.S. Dept. of Commerce.

†

2000

15.3 10.2 19.7 9.9 13.9 21.1 15.8 16.8 10.9 11.9 17.3 14.6 11.3 9.8 12.5 11.5 10.8 25.8 13.4 10.7 13.5 14.2 14.6 9.8 17.7 15.2

150 154 344 103 1,021 435 3,056 1,084 71 1,248 641 433 1,047 77 480 81 615 4,748 281 52 814 792 250 406 76 39,804

16.8 9.1 16.8 8.4 12.2 24.2 16.3 13.6 11.3 11.2 18.9 12.7 8.7 7.4 12.1 11.0 10.9 22.9 12.5 8.6 11.6 13.5 14.1 7.6 15.7 14.2

115 138 201 107 773 339 2,176 883 40 1,123 574 360 1,218 105 550 81 673 3,569 156 54 996 557 249 321 58 34,719

14.0 8.5 16.5 9.9 10.0 22.2 12.1 13.8 6.3 10.3 18.6 12.4 10.1 11.1 16.2 11.6 13.7 21.1 9.0 9.5 15.7 11.4 13.8 6.7 12.5 13.9

56

CH A P T E R 2: Describing Data Sets

3. The following table gives the 2000 and 2002 populations of some of the largest counties in the United States. Twenty-Five Largest Counties, by Population, 2000–2002 2002 Population

2000 Population

2002 Population

2000 Population

Los Angeles, CA

9,806,577

9,519,330

Broward, FL

1,709,118

1,623,018

Cook, IL

5,377,507

5,376,741

Riverside, CA

1,699,112

1,545,387

Country

County

Harris, TX

3,557,055

3,400,578

Santa Clara, CA

1,683,505

1,682,585

Maricopa, AZ

3,303,876

3,072,149

New York, NY

1,546,856

1,537,195

Orange, CA

2,938,507

2,846,289

Tarrant, TX

1,527,366

1,446,219

San Diego, CA

2,906,660

2,813,833

Clark, NV

1,522,164

1,375,738

Kings, NY

2,488,194

2,465,326

Philadelphia, PA

1,492,231

1,517,550

Miami-Dade, FL

2,332,599

2,253,362

Middlesex, MA

1,474,160

1,465,396

Dallas, TX

2,283,953

2,218,899

Alameda, CA

1,472,310

1,443,741

Queens, NY

2,237,815

2,229,379

Suffolk, NY

1,458,655

1,419,369

Wayne, MI

2,045,540

2,061,162

Bexar, TX

1,446,333

1,392,927

San Bemardino, CA

1,816,072

1,709,434

Cuyahoga,OH

1,379,049

1,393,845

King, WA

1,759,604

1,737,032

Source: Bureau of the Census. U.S. Dept of Commerce.

(a) Represent these data in a scatter diagram. (b) What conclusions can be drawn? 4. The following table gives the number of days in each year from 1993 to 2002 that did not meet acceptable air quality standards in a selection of U.S. metropolitan areas. Air Quality of Selected U.S. Metropolitan Areas, 1993–2002 Metropolitan statistical area Atlanta, GA Bakersfield, CA Baltimore, MD Boston, MA–NH Chicago, IL Dallas, TX Denver, CO Detroit, MI El Paso, TX Fresno, CA Houston, TX

1993

1994

1995

1996

1997

1998

1999

2000

2001

36 97 48 2 4 12 6 5 7 59 27

15 105 40 6 13 24 3 11 6 55 41

36 107 36 7 24 29 5 14 3 61 66

28 110 28 4 7 10 2 13 6 70 28

33 58 30 7 10 27 0 11 2 75 47

52 78 51 8 12 33 9 17 6 67 38

67 144 40 10 19 25 5 20 5 133 52

34 132 19 1 2 22 3 15 4 131 42

18 125 32 12 22 16 8 27 9 138 29

2002 24 152 42 16 21 15 8 26 13 152 23 (Continued )

2.5 Sets of Paired Data

(Continued ) Metropolitan statistical area

1993

1994

1995

1996

1997

1998

1999

2000

2001

Las Vegas, NV–AZ Los Angeles–Long Beach, CA Miami, FL Minneapolis–St. Paul, MN–WI New Haven–Meriden, CT New York, NY Orange County, CA Philadelphia, PA–NJ Phoenix–Mesa, AZ Pittsburgh, PA Riverside–San Bernardino, CA Sacramento, CA St. Louis, MO–IL Salt Lake City–Ogden, UT San Diego, CA San Francisco, CA Seattle–Bellevue–Everett, WA Ventura, CA Washington, DC–MD–VA–WV

3 134 6 0 12 11 25 62 14 14 168 20 9 5 59 0 0 43 52

3 139 1 2 13 16 15 37 10 22 150 37 33 17 46 0 3 63 22

3 113 2 5 14 21 9 38 22 27 125 41 38 5 48 2 2 66 32

14 94 1 0 8 14 9 38 15 12 118 44 23 14 31 0 6 62 18

4 60 3 0 19 23 3 38 12 21 107 17 15 2 14 0 1 45 30

5 56 8 1 9 18 6 37 14 39 96 29 24 19 33 0 3 29 47

8 56 7 1 19 25 14 32 10 40 123 69 31 8 33 10 6 24 39

2 87 2 2 9 19 31 22 10 29 145 45 18 15 31 4 7 31 11

1 88 1 2 15 19 31 29 8 52 155 49 17 15 31 12 3 25 22

2002 6 80 1 1 25 31 19 33 8 53 145 69 34 18 20 17 6 11 34

Note: Data indicate the number of days metropolitan statistical areas failed to meet acceptable air quality standards. All figures were revised based on new standards set in 1998. Includes fine particles less than or equal to 2.5 mm in diameter. Source: U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards.

(a) Draw a scatter diagram relating the 2000 and 2002 entries for each city. (b) Do higher values in 2002 tend to go with higher values in 2000? 5. The following data relate the attention span (in minutes) to a score on an IQ examination of 18 preschool-age children. Attention span 2.0 3.0 4.4 5.2 4.9 6.1

IQ score 82 88 86 94 90 99

Attention span 6.3 5.4 6.6 7.0 6.5 7.2

IQ score 105 108 112 116 122 110

Attention span 5.5 3.6 5.4 3.8 2.7 2.2

IQ score 118 128 128 130 140 142

(a) Draw a scatter diagram. (b) Give a plausible inference concerning the relation of attention span to IQ score.

57

58

CH A P T E R 2: Describing Data Sets

6. The following data relate prime lending rates and the corresponding inﬂation rate during 8 years in the 1970s. Inﬂation rate

Prime lending rate

3.3 6.2 11.0 9.1

5.2 8.0 10.8 7.9

Inﬂation rate

Prime lending rate

5.8 6.5 7.6

6.8 6.9 9.0

(a) Draw a scatter diagram. (b) Fit a straight line drawn “by hand” to the data pairs. (c) Using your straight line, predict the prime lending rate in a year whose inﬂation rate is 7.2 percent. 7. A random group of 12 high school juniors were asked to estimate the average number of hours they study each week. The grade point averages of these students were then determined, with the resulting data being as given in the following. Use it to represent these data in a scatter diagram. Hours reported working and GPA Hours 6 14 3 22 9

GPA

Hours

GPA

2.8 3.2 3.1 3.6 3.0

11 12 5 24 15

3.3 3.4 2.7 3.8 3.0

8. Problem 7 of Sec. 2.4 gives the scores of the ﬁrst 25 Super Bowl football games. For each game, let y denote the score of the winning team, and let x denote the number of points by which that team won. Draw a scatter diagram relating x and y. Do high values of one tend to go with high values of the other?

2.6 SOME HISTORICAL COMMENTS Probably the ﬁrst recorded instance of statistical graphics—that is, the representation of data by tables or graphs—was Sir Edmund Halley’s graphical analysis of barometric pressure as a function of altitude, published in 1686. Using the rectangular coordinate system introduced by the French scientist René Descartes in his study of analytic geometry, Halley plotted a scatter diagram and was then able to ﬁt a curve to the plotted data.

Key Terms

In spite of Halley’s demonstrated success with graphical plotting, almost all the applied scientists until the latter part of the 18th century emphasized tables rather than graphs in presenting their data. Indeed, it was not until 1786, when William Playfair invented the bar graph to represent a frequency table, that graphs began to be regularly employed. In 1801 Playfair invented the pie chart and a short time later originated the use of histograms to display data.

(Princeton University)

The use of graphs to represent continuous data—that is, data in which all the values are distinct—did not regularly appear until the 1830s. In 1833 the Frenchman A. M. Guerry applied the bar chart form to continuous crime data, by ﬁrst breaking up the data into classes, to produce a histogram. Systematic development of the histogram was carried out by the Belgian statistician and social scientist Adolphe Quetelet about 1846. Quetelet and his students demonstrated the usefulness of graphical analysis in their development of the social sciences. In doing so, Quetelet popularized the practice, widely followed today, of initiating a research study by ﬁrst gathering and presenting numerical data. Indeed, along with the additional steps of summarizing the data and then utilizing the methods of statistical inference to draw conclusions, this has become the accepted paradigm for research in all ﬁelds connected with the social sciences. It has also become an important technique in other ﬁelds, such as medical research (the testing of new drugs and therapies), as well as in such traditionally nonnumerical ﬁelds as literature (in deciding authorship) and history (particularly as developed by the French historian Fernand Braudel). John Tukey

The term histogram was ﬁrst used by Karl Pearson in his 1895 lectures on statistical graphics. The stem-and-leaf plot, which is a variant of the histogram, was introduced by the U.S. statistician John Tukey in 1970. In the words of Tukey, “Whereas a histogram uses a nonquantitative mark to indicate a data value, clearly the best type of mark is a digit.”

KEY TERMS Frequency: The number of times that a given value occurs in a data set. Frequency table: A table that presents, for a given set of data, each distinct data value along with its frequency. Line graph: A graph of a frequency table. The abscissa speciﬁes a data value, and the frequency of occurrence of that value is indicated by the height of a vertical line. Bar chart (or bar graph): Similar to a line graph, except now the frequency of a data value is indicated by the height of a bar. Frequency polygon: A plot of the distinct data values and their frequencies that connects the plotted points by straight lines.

59

60

CH A P T E R 2: Describing Data Sets

Symmetric data set: A data set is symmetric about a given value x0 if the frequencies of the data values x0 − c and x0 + c are the same for all values of c. Relative frequency: The frequency of a data value divided by the number of pieces of data in the set. Pie chart: A chart that indicates relative frequencies by slicing up a circle into distinct sectors. Histogram: A graph in which the data are divided into class intervals, whose frequencies are shown in a bar graph. Relative frequency histogram: A histogram that plots relative frequencies for each data value in the set. Stem-and-leaf plot: Similar to a histogram except that the frequency is indicated by stringing together the last digits (the leaves) of the data. Scatter diagram: A two-dimensional plot of a data set of paired values.

SUMMARY This chapter presented various ways to graphically represent data sets. For instance, consider the following set of 13 data values: 1, 2, 3, 1, 4, 2, 6, 2, 4, 3, 5, 4, 2 These values can be represented in a frequency table, which lists each value and the number of times it occurs in the data, as follows: A Frequency Table Value Frequency Value Frequency 1 2 3

2 4 2

4 5 6

3 1 1

The data also can be graphically pictured by either a line graph or a bar chart. Sometimes the frequencies of the different data values are plotted on a graph, and then the resulting points are connected by straight lines. This gives rise to a frequency polygon. When there are a large number of data values, often we break them up into class intervals. A bar chart plot relating each class interval to the number of data values falling in the interval is called a histogram. The y axis of this plot can represent either the class frequency (that is, the number of data values in the interval) or the proportion of all the data that lies in the class. In the former case we call the plot a frequency histogram and in the latter case a relative frequency histogram.

Summary

A line graph.

A bar graph.

A frequency polygon.

Consider this data set: 41, 38, 44, 47, 33, 35, 55, 52, 41, 66, 64, 50, 49, 56, 55, 48, 52, 63, 59, 57, 75, 63, 38, 37 Using the ﬁve class intervals 30–40, 40–50, 50–60, 60–70, 70–80

61

62

CH A P T E R 2: Describing Data Sets

A histogram.

along with the left-end inclusion convention (which signiﬁes that the interval contains all points greater than or equal to its left-end member and less than its right-end member), we have the histogram above to represent this data set. Data sets can also be graphically displayed in a stem-and-leaf plot. The following stem-and-leaf plot is for the preceding data set. 7 6 5 4 3

5 3,3,4,6 0,2,2,5,5,6,7,9 1,1,4,7,8,9 3,5,7,8,8

A stem-and-leaf plot.

Often data come in pairs. That is, for each element of the data set there is an x value and a y value. A plot of the x and y values is called a scatter diagram. A scatter diagram can be quite useful in ascertaining such things as whether high x values appear to go along with high y values, or whether high x values tend to go along with low y values, or whether there appears to be no particular association between the x and y values of a pair. The following data set of pairs i 1 8 xi yi 14

2 12 10

3 7 17

4 15 9

5 5 13

6 12 8

7 10 12

8 22 6

Review Problems

A scatter diagram.

is represented in the scatter diagram above. The diagram indicates that high values of one member of the pair appear to be generally associated with low values of the other member of the pair. Using these graphical tools, often we can communicate pertinent features of a data set at a glance. As a result, we can learn things about the data that are not immediately evident in the raw numbers themselves. The choice of which display to use depends on such things as the size of the data set, the type of data, and the number of distinct values.

REVIEW PROBLEMS 1. The following data are the blood types of 50 volunteers at a blood plasma donation clinic: O A O AB A A O O B A O A AB B O O O A B A A O A A O B A O AB A O O A B A A A O B O O A O A B O AB A O B (a) Represent these data in a frequency table. (b) Represent them in a relative frequency table. (c) Represent them in a pie chart.

63

64

CH A P T E R 2: Describing Data Sets

2. The following is a sample of prices, rounded to the nearest cent, charged per gallon of standard gasoline in the San Francisco Bay area in May 1991: 121, 119, 117, 121, 120, 120, 118, 124, 123, 139, 120, 115, 117, 121, 123, 120, 123, 118, 117, 122, 122, 119 (a) Construct a frequency histogram for this data set. (b) Construct a frequency polygon. (c) Construct a stem-and-leaf plot. (d) Does any data value seem out of the ordinary? If so, explain why. 3. The following frequency table presents the number of female suicides that took place in eight German states over 14 years. Number of suicides per year 0 Frequency

1

2

3

4

5 6 7 8 9 10

9 19 17 20 15 11 8 2 3 5

3

Thus, for instance, there were a total of 20 cases in which states had 3 suicides in a year. (a) How many suicides were reported over the 14 years? (b) Represent the above data in a histogram. 4. The following table gives the 1991 crime rate (per 100,000 population) in each state. Use it to construct a (a) Frequency histogram of the total violent crime rates in the northeastern states (b) Relative frequency histogram of the total property crime rates in the southern states (c) Stem-and-leaf plot of the murder rates in the western states (d) Stem-and-leaf plot of the burglary rates in the midwestern states. Violent crime Region, Division, and State

Total Total Murder

United States Northeast New England Maine New Hampshire Vermont Massachusetts

5,898 5,155 4,950 3,768 3,448 3,955 5,332

758 752 532 132 119 117 736

9.8 8.4 4.1 1.2 3.6 2.1 4.2

Property crime

Forcible Aggravated Larceny— Motor rape Robbery assault Total Burglary theft vehicle theft 42 29 30 22 30 31 32

273 352 159 23 33 12 195

433 363 338 86 53 72 505

5,140 4,403 4,419 3,636 3,329 3,838 4,586

1,252 1,010 1,103 903 735 1,020 1,167

3,229 2,598 2,600 2,570 2,373 2,674 2,501

659 795 716 163 220 144 919 (Continued)

Review Problems

Violent crime Region, Division, and State

Total Total Murder

Rhode Island 5,039 462 Connecticut 5,364 540 Middle Atlantic 5,227 829 New York 6,245 1,164 New Jersey 5,431 635 Pennsylvania 3,559 450 Midwest 5,257 631 East north central 5,482 704 Ohio 5,033 562 Indiana 4,818 505 Illinois 6,132 1,039 Michigan 6,138 803 Wisconsin 4,466 277 West north central 4,722 457 Minnesota 4,496 316 Iowa 4,134 303 Missouri 5,416 763 North Dakota 2,794 65 South Dakota 3,079 182 Nebraska 4,354 335 Kansas 5,534 500 South 6,417 798 South Atlantic 6,585 851 Delaware 5,869 714 Maryland 6,209 956 District of Columbia 10,768 2,453 Virginia 4,607 373 West Virginia 2,663 191 North Carolina 5,889 658 South Carolina 6,179 973 Georgia 6,493 738 Florida 8,547 1,184 East south central 4,687 631 Kentucky 3,358 438 Tennessee 5,367 726 Alabama 5,366 844 Mississippi 4,221 389 West south central 7,118 806

3.7 5.7 9.9 14.2 5.2 6.3 7.8 8.9 7.2 7.5 11.3 10.8 4.8 5.4 3.0 2.0 10.5 1.1 1.7 3.3 6.1 12.1 11.4 5.4 11.7 80.6 9.3 6.2 11.4 11.3 12.8 9.4 10.4 6.8 11.0 11.5 12.8 14.2

65

Property crime

Forcible Aggravated Larceny— Motor rape Robbery assault Total Burglary theft vehicle theft 31 29 29 28 29 29 45 50 53 41 40 79 25 34 40 21 34 18 40 28 45 45 44 86 46 36 30 23 35 59 42 52 41 35 46 36 46 50

123 224 419 622 293 194 223 263 215 116 456 243 119 129 98 45 251 8 19 54 138 252 286 215 407 1,216 138 43 178 171 268 400 149 83 213 153 116 254

304 280 372 499 307 221 355 383 287 340 532 470 128 288 175 235 467 38 122 249 310 489 510 408 492 1,121 196 119 434 731 415 723 430 313 456 644 214 488

4,577 4,824 4,398 5,081 4,797 3,109 4,626 4,777 4,471 4,312 5,093 5,335 4,189 4,265 4,180 3,831 4,653 2,729 2,897 4,020 5,035 5,618 5,734 5,155 5,253 8,315 4,234 2,472 5,230 5,207 5,755 7,363 4,056 2,920 4,641 4,521 3,832 6,312

1,127 1,191 978 1,132 1,016 720 1,037 1,056 1,055 977 1,120 1,186 752 991 854 832 1,253 373 590 727 1,307 1,498 1,508 1,128 1,158 2,074 783 667 1,692 1,455 1,515 2,006 1,196 797 1,365 1,269 1,332 1,653

2,656 2,838 2,598 2,944 2,855 1,907 3,082 3,151 2,916 2,871 3,318 3,469 3,001 2,918 2,963 2,828 2,841 2,229 2,192 3,080 3,377 3,518 3,665 3,652 3,365 4,880 3,113 1,631 3,239 3,365 3,629 4,573 2,465 1,909 2,662 2,889 2,213 3,871

794 796 823 1,004 926 482 507 570 500 465 655 680 436 356 363 171 558 127 115 213 351 603 561 375 731 1,360 339 175 299 387 611 784 395 215 614 363 286 788 (Continued)

66

CH A P T E R 2: Describing Data Sets

(Continued) Violent crime Region, Division, and State

Total Total Murder

Arkansas Louisiana Oklahoma Texas West Mountain Montana Idaho Wyoming Colorado New Mexico Arizona Utah Nevada Pacific Washington Oregon California Alaska Hawaii

5,175 593 6,425 951 5,669 584 7,819 840 6,478 841 6,125 544 3,648 140 4,196 290 4,389 310 6,074 559 6,679 835 7,406 671 5,608 287 6,299 677 6,602 945 6,304 523 5,755 506 6,773 1,090 5,702 614 5,970 242

11.1 16.9 7.2 15.3 9.6 6.5 2.6 1.8 3.3 5.9 10.5 7.8 2.9 11.8 10.7 4.2 4.6 12.7 7.4 4.0

Property crime

Forcible Aggravated Larceny— Motor rape Robbery assault Total Burglary theft vehicle theft 45 41 51 53 46 44 20 29 26 47 52 42 46 66 47 70 53 42 92 33

136 279 129 286 287 122 19 21 17 107 120 166 55 312 345 146 150 411 113 87

402 614 397 485 498 371 99 239 264 399 652 455 183 287 542 303 298 624 402 118

4,582 5,473 5,085 6,979 5,637 5,581 3,508 3,905 4,079 5,515 5,845 6,735 5,321 5,622 5,656 5,781 5,249 5,683 5,088 5,729

1,227 1,412 1,478 1,802 1,324 1,247 524 826 692 1,158 1,723 1,607 840 1,404 1,351 1,235 1,176 1,398 979 1,234

3,014 3,489 3,050 4,232 3,522 3,843 2,778 2,901 3,232 3,930 3,775 4,266 4,240 3,565 3,409 4,102 3,598 3,246 3,575 4,158

341 573 557 944 791 491 206 178 155 426 346 861 241 652 896 444 474 1,039 534 336

Source: U.S. Federal Bureau of Investigation, Crime in the United States, annual.

5. Construct a frequency table for a data set of 10 values that is symmetric and has (a) 5 distinct values and (b) 4 distinct values. (c) About what values are the data sets in parts (a) and (b) symmetric? 6. The following are the estimated oil reserves, in billions of barrels, for four regions in the western hemisphere. Represent the data in a pie chart. United States South America Canada Mexico

38.7 22.6 8.8 60.0

7. The following pie chart represents the percentages of the world’s 2006 total military spending by countries and regions of the world. Use it to estimate the percentages of all military expenditures spent by (a) the United States, and (b) China.

Review Problems

Source: Stockholm International Peace Research Institude Yearbook 2007.

8. The following data refer to the ages (to the nearest year) at which patients died at a large inner-city (nonbirthing) hospital: 1, 1, 1, 1, 3, 3, 4, 9, 17, 18, 19, 20, 20, 22, 24, 26, 28, 34, 45, 52, 56, 59, 63, 66, 68, 68, 69, 70, 74, 77, 81, 90 (a) (b) (c) (d)

Represent this data set in a histogram. Represent it in a frequency polygon. Represent it in a cumulative frequency polygon. Represent it in a stem-and-leaf plot.

Problems 9 to 11 refer to the last 50 student entries in App. A. 9. (a) Draw a histogram of the weights of these students. (b) Comment on this histogram. 10. Draw a scatter diagram relating weight and cholesterol level. Comment on what the scatter diagram indicates.

67

68

CH A P T E R 2: Describing Data Sets

11. Draw a scatter diagram relating weight and blood pressure. What does this diagram indicate? Problems 12 and 13 refer to the following table concerning the mathematics and verbal SAT scores of a graduating class of high school seniors.

Student 1 2 3 4 5 6 7

Verbal Mathematics Verbal Mathematics score score Student score score 520 605 528 720 630 504 530

505 575 672 780 606 488 475

8 9 10 11 12 13 14

620 604 720 490 524 646 690

576 622 704 458 552 665 550

12. Draw side-by-side stem-and-leaf plots of the student scores on the mathematics and verbal SAT examinations. Did the students, as a group, perform better on one examination? If so, which one? 13. Draw a scatter diagram of student scores on the two examinations. Do high scores on one tend to go along with high scores on the other? 14. The following table gives information about the age of the population in both the United States and Mexico. Proportion of population (percent) Age, years 0–9 10–19 20–29 30–39 40–49 50–59 60–69 70–79 Over 80

Mexico 32.5 24 14.5 11 7.5 4.5 3.5 1.5 1

United States 17.5 20 14.5 12 12.5 10.5 7 4 2

(a) What percentage of the Mexican population is less than 30 years old? (b) What percentage of the U.S. population is less than 30 years old? (c) Draw two relative frequency polygons on the same graph. Use different colors for Mexican and for U.S. data. (d) In general, how do the age distributions compare for the two countries?

Review Problems

15. The following data relate to the normal monthly and annual precipitation (in inches) for various cities. Normal Monthly and Annual Precipitation in Selected Cities State

City

Jan.

Feb.

Mar.

Apr.

May

June

July

Aug.

Sept.

Oct.

Nov.

Dec.

Annual

AL

Mobile

4.59

4.91

6.48

5.35

5.46

5.07

7.74

6.75

6.56

2.62

3.67

5.44

64.64

AK

Juneau

3.69

3.74

3.34

2.92

3.41

2.98

4.13

5.02

6.40

7.71

5.15

4.66

53.15

AZ

Phoenix

0.73

0.59

0.81

0.27

0.14

0.17

0.74

1.02

0.64

0.63

0.54

0.83

7.11

AR

Little Rock

3.91

3.83

4.69

5.41

5.29

3.67

3.63

3.07

4.26

2.84

4.37

4.23

49.20

CA

Los Angeles

3.06

2.49

1.76

0.93

0.14

0.04

0.01

0.10

0.15

0.26

1.52

1.62

12.08

Sacramento

4.03

2.88

2.06

1.31

0.33

0.11

0.05

0.07

0.27

0.86

2.23

2.90

17.10

San Diego

2.11

1.43

1.60

0.78

0.24

0.06

0.01

0.11

0.19

0.33

1.10

1.36

9.32

San Francisco

4.65

3.23

2.64

1.53

0.32

0.11

0.03

0.05

0.19

1.06

2.35

3.55

19.71

CO

Denver

0.51

0.69

1.21

1.81

2.47

1.58

1.93

1.53

1.23

0.98

0.82

0.55

15.31

CT

Hartford

3.53

3.19

4.15

4.02

3.37

3.38

3.09

4.00

3.94

3.51

4.05

4.16

44.39

DE

Wilmington

3.11

2.99

3.87

3.39

3.23

3.51

3.90

4.03

3.59

2.89

3.33

3.54

41.38

DC

Washington

2.76

2.62

3.46

2.93

3.48

3.35

3.88

4.40

3.22

2.90

2.82

3.18

39.00

FL

Jacksonville

3.07

3.48

3.72

3.32

4.91

5.37

6.54

7.15

7.26

3.41

1.94

2.59

52.76

Miami

2.08

2.05

1.89

3.07

6.53

9.15

5.98

7.02

8.07

7.14

2.71

1.86

57.55

GA

Atlanta

4.91

4.43

5.91

4.43

4.02

3.41

4.73

3.41

3.17

2.53

3.43

4.23

48.61

HI

Honolulu

3.79

2.72

3.48

1.49

1.21

0.49

0.54

0.60

0.62

1.88

3.22

3.43

23.47

ID

Boise

1.64

1.07

1.03

1.19

1.21

0.95

0.26

0.40

0.58

0.75

1.29

1.34

11.71

IL

Chicago

1.60

1.31

2.59

3.66

3.15

4.08

3.63

3.53

3.35

2.28

2.06

2.10

33.34

Peoria

1.60

1.41

2.86

3.81

3.84

3.88

3.99

3.39

3.63

2.51

1.96

2.01

34.89

IN

Indianapolis

2.65

2.46

3.61

3.68

3.66

3.99

4.32

3.46

2.74

2.51

3.04

3.00

39.12

IA

Des Moines

1.01

1.12

2.20

3.21

3.96

4.18

3.22

4.11

3.09

2.16

1.52

1.05

30.83

KS

Wichita

0.68

0.85

2.01

2.30

3.91

4.06

3.62

2.80

3.45

2.47

1.47

0.99

28.61

KY

Louisville

3.38

3.23

4.73

4.11

4.15

3.60

4.10

3.31

3.35

2.63

3.49

3.48

43.56

LA

New Orleans

4.97

5.23

4.73

4.50

5.07

4.63

6.73

6.02

5.87

2.66

4.06

5.27

59.74

Source: U.S. National Oceanic and Atmospheric Administration, Climatography of the United States, September 1982.

(a) Represent the normal precipitation amounts for April in a stemand-leaf plot. (b) Represent the annual amounts in a histogram. (c) Draw a scatter diagram relating the April amount to the annual amount. 16. A data value that is far away from the other values is called an outlier. In the following data sets, specify which, if any, of the data values are outliers. (a) 14, 22, 17, 5, 18, 22, 10, −17, 25, 28, 33, 12 (b) 5, 2, 13, 16, 9, 12, 7, 10, 54, 22, 18, 15, 12 (c) 18, 52, 14, 20, 24, 27, 43, 17, 25, 28, 3, 22, 6

69

70

CH A P T E R 2: Describing Data Sets

17. The following table presents data on the number of cars imported from Japan and from Germany in the years 1970 to 2002. New Passenger Cars Imported Into the United States Japan 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986

381, 338 703, 672 697, 788 624, 805 791, 791 695, 573 1, 128, 936 1, 341, 530 1, 563, 047 1, 617, 328 1, 991, 502 1, 911, 525 1, 801, 185 1, 871, 192 1, 948, 714 2, 527, 467 2, 618, 711

Japan

Germany 674, 945 770, 807 676, 967 677, 465 619, 757 370, 012 349, 804 423, 492 416, 231 495, 565 338, 711 234, 052 259, 385 239, 807 335, 032 473, 110 451, 699

1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

2, 417, 509 2, 123, 051 2, 051, 525 1, 867, 794 1, 762, 347 1, 598, 919 1, 501, 953 1, 488, 159 1, 114, 360 1, 190, 896 1, 387, 812 1, 456, 081 1, 707, 277 1, 839, 093 1, 790, 346 2, 046, 902

Germany 377, 542 264, 249 216, 881 245, 286 171, 097 205, 248 180, 383 178, 774 204, 932 234, 909 300, 489 373, 330 461, 061 488, 323 494, 131 574, 455

Source: Bureau of the Census, Foreign Trade Division.

(a) What conclusions can you draw concerning the yearly number of Japanese and of German cars imported into the United States since 1990? (b) Present a scatter diagram relating Japanese and German car imports since 1990.

CHAPTER 3

Using Statistics to Summarize Data Sets I do hate averages. There is no greater mistake than to call arithmetic an exact science. There are permutations and aberrations discernible to minds entirely noble like mine; subtle variations which ordinary accountants fail to discover, hidden laws of numbers which it requires a mind like mine to perceive. For instance if you average numbers from the bottom up and then again from the top down, the result is always different. A letter to the Mathematical Gazette (a 19th-century British mathematical journal)

The way to make sense out of raw data is to compare and contrast, to understand differences. Gregory Bateson (in Steps to an Ecology of the Mind)

CONTENTS 3.1

Introduction .........................................................................

72

3.2

Sample Mean .......................................................................

73

Problems .............................................................................

79

Sample Median ....................................................................

83

Problems .............................................................................

86

Sample Mode .......................................................................

97

Problems .............................................................................

98

Sample Variance and Sample Standard Deviation .....................

99

3.3 3.4 3.5

Problems ............................................................................. 105 3.6

Normal Data Sets and the Empirical Rule................................. 109 Problems ............................................................................. 114

Introductory Statistics, DOI: 10.1016/B978-0-12-374388-6.00003-X © 2010, Elsevier Inc. All rights reserved.

71

72

CH A P T E R 3: Using Statistics to Summarize Data Sets

3.7

Sample Correlation Coefﬁcient ............................................... 120 Problems ............................................................................. 128

Key Terms .................................................................................. 134 Summary .................................................................................... 136 Review Problems ......................................................................... 138 Our objective in this chapter is to develop measures that can be used to summarize a data set. These measures, formally called statistics, are quantities whose values are determined by the data. We study the sample mean, sample median, and sample mode. These are all statistics that measure the center or middle value of a data set. Statistics that indicate the amount of variation in the data set are also considered. We learn about what it means for a data set to be normal, and we present an empirical rule concerning such sets. We also consider data sets consisting of paired values, and we present a statistic that measures the degree to which a scatter diagram of paired values can be approximated by a straight line.

3.1 INTRODUCTION Modern-day experiments often track certain characteristics of thousands of individuals over time. For instance, in an attempt to learn about the health consequences of certain common practices, the medical statisticians R. Doll and A. B. Hill sent questionnaires in 1951 to all doctors in the United Kingdom and received 40,000 replies. Their questionnaire dealt with age, eating habits, exercise habits, and smoking habits. These doctors were then monitored for 10 years, and the causes of death of those who died were determined. As one can imagine, this study resulted in huge sets of data. For instance, even if we just focus on one component of the study at a single moment of time, such as the doctors’ ages in 1951, the resulting data set of 40,000 values is vast. To obtain a feel for such a large data set, it is often necessary to summarize it by some suitably chosen measures. In this chapter, we introduce different statistics that can be used to summarize certain features of data sets. To begin, suppose that we have in our possession sample data from some underlying population. Now, whereas in Chap. 2 we showed how to describe and portray data sets in their entirety, here we will be concerned with determining certain summary measures about the data. These summary measures are called statistics, where by a statistic we mean any numerical quantity whose value is determined by the data. Deﬁnition Numerical quantities computed from a data set are called statistics. We will be concerned with statistics that describe the central tendency of the data set; that is, they describe the center of the set of data values. Three different

3.2 Sample Mean

statistics for describing this—the sample mean, sample median, and sample mode—will be presented in Secs. 3.2, 3.3, and 3.4, respectively. Once we have some idea of the center of a data set, the question naturally arises as to how much variation there is. That is, are most of the values close to the center, or do they vary widely about the center? In Sec. 3.5 we will discuss the sample variance and sample standard deviation, which are statistics designed to measure such variation. In Sec. 3.6 we introduce the concept of a normal data set, which is a data set having a bell-shaped histogram. For data sets that are close to being normal, we present a rule that can be used to approximate the proportion of the data that is within a speciﬁed number of sample standard deviations from the sample mean. In the ﬁrst six sections of this chapter, we concern ourselves with data sets where each datum is a single value. However, in Sec. 3.7 we deal with paired data. That is, each data point will consist of an x value and a y value. For instance, the x value might represent the average number of cigarettes that an individual smoked per day, and the y value could be the age at which that individual died. We introduce a statistic called the sample correlation coefﬁcient whose value indicates the degree to which data points having large x values also have large y values and correspondingly the degree to which those having small x values also have small y values. The Doll–Hill study yielded the result that only about 1 in 1000 nonsmoking doctors died of lung cancer. For heavy smokers the ﬁgure was 1 in 8. In addition, death rates from heart attacks were 50 percent higher for smokers.

3.2 SAMPLE MEAN Suppose we have a sample of n data points whose values we designate by x1 , x2 , . . . , xn . One statistic for indicating the center of this data set is the sample mean, deﬁned to equal the arithmetic average of the data values. Deﬁnition The sample mean, which we designate by x (pronounced “x bar”), is deﬁned by n x 1 + x 2 + · · · + xn xi x = i=1 = n n ■

Example 3.1 The average fuel efﬁciencies, in miles per gallon, of cars sold in the United States in the years 1999 to 2003 were 28.2, 28.3, 28.4, 28.5, 29.0 Find the sample mean of this set of data.

73

74

CH A P T E R 3: Using Statistics to Summarize Data Sets

Solution The sample mean x is the average of the ﬁve data values. Thus, x¯ =

142.4 28.2 + 28.3 + 28.4 + 28.5 + 29.0 = = 28.48 5 5

Note from this example that whereas the sample mean is the average of all the data values, it need not itself be one of them. ■ Consider again the data set x1 , x2 , . . . , xn . If each data value is increased by a constant amount c, then this causes the sample mean also to be increased by c. Mathematically, we can express this by saying that if yi = xi + c

i = 1, . . . , n

then y =x+c where y and x are the sample means of the yi and the xi , respectively. Therefore, when it is convenient, we can compute x by ﬁrst adding c to all the data values, then computing the sample mean y of the new data, and ﬁnally subtracting c from y to obtain x. Since it is sometimes a lot easier to work with the transformed rather than the original data, this can greatly simplify the computation of x. Our next example illustrates this point. ■

Example 3.2 The winning scores in the U.S. Masters Golf Tournament in the years from 1981 to 1990 were as follows: 280, 284, 280, 277, 282, 279, 285, 281, 283, 278 Find the sample mean of these winning scores. Solution Rather than directly adding the preceding numbers, ﬁrst we subtract 280 from (that is, add c = −280 to) each one to obtain the following transformed data: 0, 4, 0, −3, 2, −1, 5, 1, 3, −2 The sample mean of these transformed data, call it y, is y=

0+4+0−3+2−1+5+1+3−2 9 = 10 10

Adding 280 to y shows that the sample mean of the original data is x = 280.9

■

3.2 Sample Mean

If each data value is multiplied by c, then so is the sample mean. That is, if yi = cxi

i = 1, . . . , n

then y = cx For instance, suppose that the sample mean of the height of a collection of individuals is 5.0 feet. Suppose that we now want to change the unit of measurement from feet to inches. Then since each new data value is the old value multiplied by 12, it follows that the sample mean of the new data is 12 · 5 = 60. That is, the sample mean is 60 inches. Our next example considers the computation of the sample mean when the data are arranged in a frequency table. ■

Example 3.3 The number of suits sold daily by a women’s boutique for the past 6 days has been arranged in the following frequency table: Value Frequency 3 4 5

2 1 3

What is the sample mean? Solution Since the original data set consists of the 6 values 3, 3, 4, 5, 5, 5 it follows that the sample mean is 3+3+4+5+5+5 6 3×2+4×1+5×3 = 6 25 = 6

x=

That is, the sample mean of the number of suits sold daily is 4.25.

■

75

76

CH A P T E R 3: Using Statistics to Summarize Data Sets

In Example 3.3 we have seen that when the data are arranged in a frequency table, the sample mean can be expressed as the sum of the products of the distinct values and their frequencies, all divided by the size of the data set. This result holds in general. To see this, suppose the data are given in a frequency table that lists k distinct values x1 , x2 , . . . , xk with respective frequencies f1 , f2 , . . . , fk . It follows that the data set consists of n observations, where n = ki=1 fi and where the value xi appears fi times for i = 1, 2, . . . , k. Hence, the sample mean for this data set is x1 + · · · + x1 + x2 + · · · + x2 + · · · + xk + · · · + xk n f1 x1 + f2 x2 + · · · + fk xk = n

x=

(3.1)

Now, if w1 , w2 , . . . , wk are nonnegative numbers that sum to 1, then w1 x1 + w2 x2 + · · · + wk xk is said to be a weighted average of the values x1 , x2 , . . . , xk with wi being the weight of xi . For instance, suppose that k = 2. Now, if w1 = w2 = 1/2, then the weighted average 1 1 w1 x1 + w2 x2 = x1 + x2 2 2 is just the ordinary average of x1 and x2 . On the other hand, if w1 = 2/3 and w2 = 1/3, then the weighted average w1 x1 + w2 x2 =

2 1 x 1 + x2 3 3

gives twice as much weight to x1 as it does to x2 . By writing Eq. (3.1) as x=

f2 fk f1 x1 + x2 + · · · + xk n n n

we see that the sample mean x is a weighted average of the set of distinct values. The weight given to the value xi is fi /n, the proportion of the data values that is equal to xi . Thus, for instance, in Example 3.3 we could have written that x=

■

2 1 3 25 ×3+ ×4+ ×5= 6 6 6 6

Example 3.4 In a paper entitled “The Effects of Helmet Use on the Severity of Head Injuries in Motorcycle Accidents” (published in the Journal of the American Statistical Association, 1992, pp. 48–56), A. Weiss analyzed a sample of 770 similar motorcycle

3.2 Sample Mean

accidents that occurred in the Los Angeles area in 1976 and 1977. Each accident was classiﬁed according to the severity of the head injury suffered by the motorcycle operator. The classiﬁcation used was as follows: Classiﬁcation of accident

Interpretation

0 1 2 3 4 5 6

No head injury Minor head injury Moderate head injury Severe, not life-threatening Severe and life-threatening Critical, survival uncertain at time of accident Fatal

In 331 of the accidents the operator wore a helmet, whereas in the other 439 accidents the operator did not. The following are frequency tables giving the severities of the accidents that occurred when the operator was wearing and was not wearing a helmet.

Classiﬁcation 0 1 2 3 4 5 6

Frequency of driver with helmet

Frequency of driver without helmet

248 58 11 3 2 8 1

227 135 33 14 3 21 6

331

439

Find the sample mean of the head severity classiﬁcations for those operators who wore helmets and for those who did not. Solution The sample mean for those wearing helmets is x=

0.248 + 1.58 + 2.11 + 3.3 + 4.2 + 5.8 + 6.1 143 = = 0.432 331 331

The sample mean for those who did not wear a helmet is x=

396 0.227 + 1.135 + 2.33 + 3.14 + 4.3 + 5.21 + 6.6 = = 0.902 439 439

77

78

CH A P T E R 3: Using Statistics to Summarize Data Sets

Therefore, the data indicate that those cyclists who were wearing a helmet suffered, on average, less severe head injuries than those who were not wearing a helmet. ■

3.2.1 Deviations Again n suppose that sample data consist of the n values x1 , . . . , xn and that x = i=1 xi /n is the sample mean. The differences between each of the data values and the sample mean are called deviations. Deﬁnition The deviations are the differences between the data values and the sample mean. The value of the ith deviation is x i − x. A useful identity is that the sum of all the deviations must equal 0. That is, n

(xi − x) = 0

i=1

That this equality is true is seen by the following argument: n i=1

(xi − x) =

n

xi −

i=1

n

x

i=1

= nx − nx =0 This equality states that the sum of the positive deviations from the sample mean must exactly balance the sum of the negative deviations. In physical terms, this means that if n weights of equal mass are placed on a (weightless) rod at the points xi , i = 1, . . . , n, then x is the point at which the rod will be in balance. This balancing point is called the center of gravity (Fig. 3.1).

Historical Perspective In the early days of sea voyages it was quite common for large portions of a ship’s cargo to be either lost or damaged due to storms. To handle this potential loss, there was a standard agreement that all those having merchandise aboard the ship would contribute to pay for the value of all lost or damaged goods. The amount of money that each of them was called upon to pay was known as havaria, and from this Latin word derives our present word average. (Typically, if there were n shippers having damages x1 , . . . , xn , then the total loss was x1 + · · · + xn and the havaria for each was (x1 + · · · + xn )/n.)

3.2 Sample Mean

FIGURE 3.1 The center of gravity of 0, 1, 2, 6, 10, 11 is (0 + 1 + 2 + 6 + 10 + 11)/6 = 30/6 = 5.

■

Example 3.5 For the data of Example 3.1, the deviations from the sample mean of 28.48 are x1 − x = 28.2 − 28.48 = −0.28 x2 − x = 28.3 − 28.48 = −0.18 x3 − x = 28.4 − 28.48 = −0.08 x4 − x = 28.5 − 28.48 = 0.02 x5 − x = 29.0 − 28.48 = 0.52 As a check, we note that the sum of the deviations is −0.28 − 0.18 − 0.08 + 0.02 + 0.52 = 0

■

PROBLEMS 1. The following data represent the scores on a statistics examination of a sample of students: 87, 63, 91, 72, 80, 77, 93, 69, 75, 79, 70, 83, 94, 75, 88 What is the sample mean? 2. The following data (from U.S. Department of Agriculture, Food Consumption, Prices, and Expenditures) give the U.S. per capita consumption (in pounds) of cheese in a sample of years. Year

1965

1975

1985

1995

2001

Per capita consumption

10.0

14.8

23.4

26.4

30.1

Find the sample mean of the given data. 3. The following data give the annual average number of inches of precipitation and the average number of days of precipitation in a sample of cities.

79

80

CH A P T E R 3: Using Statistics to Summarize Data Sets

City Albany, NY Baltimore, MD Casper, WY Denver, CO Fargo, ND Houston, TX Knoxville, TN Los Angeles, CA Miami, FL New Orleans, LA Pittsburgh, PA San Antonio, TX Wichita, KS

Average amount of precipitation

Average number of days

35.74 31.50 11.43 15.31 19.59 44.76 47.29 12.08 57.55 59.74 36.30 29.13 28.61

134 83 95 88 100 105 127 36 129 114 154 81 85

Source: National Oceanic and Atmospheric Administration.

4.

5.

6.

7.

(a) Find the sample mean of the average number of inches of precipitation. (b) Find the sample mean of the average number of days of precipitation. Consider ﬁve numbers. Suppose the mean of the ﬁrst four numbers is 14. (a) If the ﬁfth number is 24, what is the mean of all ﬁve numbers? (b) If the mean of all ﬁve numbers is 24, what is the ﬁfth number? The sample mean of the weights of the adult women of town A is larger than the sample mean of the weights of the adult women of town B. Moreover, the sample mean of the weights of the adult men of town A is larger than the sample mean of the weights of the adult men of town B. Can we conclude that the sample mean of the weights of the adults of town A is larger than the sample mean of the weights of the adults of town B? Explain your answer. Suppose that the sample mean of a set of 10 data points is x = 20. (a) If it is discovered that a data point having value 15 was incorrectly read as having value 13, what should be the revised value of x? (b) Suppose there is an additional data point whose value is 22. Will this increase or decrease the value of x? (c) Using the original data (and not the revised data in part (a)), what is the new value of x in part (b)? The following table gives the number of cases of tetanus in the 27-country European community in the years from 1996 to 2004. Find the sample mean of these 27 × 11 = 297 data values.

3.2 Sample Mean

Tetanus Number of cases

EU-27 Belgium Bulgaria Czech Republic Denmark Germany Estonia Ireland Greece Spain France Italy Cyprus Latvia Lithuania Luxembourg Hungary Malta Netherlands Austria Poland Portugal Romania Slovenia Slovakia Finland Sweden United Kingdom

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

352 3 4 0 0 17 1 0 7 45 39 105 0 4 2 1 11 1 2 0 46 23 22 5 1 1 3 9

309 1 5 2 2 11 2 0 2 45 29 103 0 3 3 0 12 1 2 0 37 16 17 5 0 0 3 8

290 0 1 1 2 7 1 1 2 32 20 119 0 1 2 0 12 1 3 0 22 24 23 3 0 2 2 9

288 1 10 0 1 8 1 1 6 38 17 91 0 3 6 0 20 0 1 0 21 25 19 5 0 8 2 4

263 1 3 1 4 8 1 1 16 29 29 98 0 2 0 0 10 1 5 0 14 15 14 9 0 : 0 2

220 3 4 3 1 8 1 3 4 23 28 63 0 1 1 0 8 1 0 0 21 15 23 2 0 : 1 6

188 1 2 0 0 : 0 0 3 21 17 69 1 0 1 0 5 2 : : 20 11 21 5 2 : 0 7

219 1 2 0 0 : 1 0 7 24 30 73 2 0 4 0 4 1 4 0 30 6 13 3 0 0 0 14

177 2 0 0 1 : 0 1 5 16 25 56 0 1 1 0 1 1 : : 25 9 11 2 0 : 0 20

98 3 2 0 0 : 0 0 7 18 17 : 0 0 4 0 3 0 : : 15 8 8 2 0 : 1 10

8. The following stem-and-leaf plot portrays the most recent 15 league bowling scores of the author of this text. Compute the sample mean. 18 17 16 15 14 13

2, 4, 7 0 1,9 2, 2, 4, 8, 8 2, 1, 5, 5

9. Find the sample mean for this data set: 1, 2, 4, 7, 10, 12

81

82

CH A P T E R 3: Using Statistics to Summarize Data Sets

Now ﬁnd the sample means for the data sets 3, 6, 12, 21, 30, 36 and 6, 7, 9, 12, 15, 17 10. Suppose that x is the sample mean of the data set consisting of the data x1 , . . . , xn . If the data are transformed according to the formula yi = axi + b

i = 1, . . . , n

what is the sample mean of the data set y1 , . . . , yn ? (In the equation, a and b are given constants.) 11. The following data give the total number of ﬁres in Ontario, Canada, in the months of 2002: 6, 13, 5, 7, 7, 3, 7, 2, 5, 6, 9, 8 Find the sample mean of this data set. 12. The following data set speciﬁes the total number of cars produced in the United States over a sample of years. The data are in units of 1000 cars. Find the sample mean of the number of cars sold annually in these years. Year

1980

1985

1990

1995

2000

2002

2006

Number sold

8010

11,653

9783

11,985

12,832

12,326

11,264

Source: Statistical Abstract of the United States, 2008.

13. One-half the values of a sample are equal to 10, and the other half are equal to 20. What is the sample mean? 14. The following is a frequency table of the ages of a sample of members of a symphony for young adults. Age value 16 17 18 19 20

Frequency 9 12 15 10 8

Find the sample mean of the given ages. 15. Half the values of a sample are equal to 10, one-sixth are equal to 20, and one-third are equal to 30. What is the sample mean? 16. There are two entrances to a parking lot. Student 1 counts the daily number of cars that pass through entrance 1, and student 2 does the same for entrance 2. Over 30 days, the data of student 1 yielded a

3.3 Sample Median

17.

18.

19.

20.

sample mean of 122, and the data of student 2 yielded a sample mean of 160. Over these 30 days, what was the daily average number of cars that entered the parking lot? A company runs two manufacturing plants. A sample of 30 engineers at plant 1 yielded a sample mean salary of $33,600. A sample of 20 engineers at plant 2 yielded a sample mean salary of $42,400. What is the sample mean salary for all 50 engineers? Suppose that we have two distinct samples of sizes n1 and n2 . If the sample mean of the ﬁrst sample is x 1 and that of the second is x 2 , what is the sample mean of the combined sample of size n1 + n2 ? Find the deviations for each of the three data sets of Prob. 9, and verify your answers by showing that in each case the sum of the deviations is 0. Calculate the deviations for the data of Prob. 14 and check that they sum to 0.

3.3 SAMPLE MEDIAN The following data represent the number of weeks after completion of a learn-todrive course that it took a sample of seven people to obtain a driver’s license: 2, 110, 5, 7, 6, 7, 3 The sample mean of this data set is x = 140/7 = 20; and so six of the seven data values are quite a bit less than the sample mean, and the seventh is much greater. This points out a weakness of the sample mean as an indicator of the center of a data set—namely, its value is greatly affected by extreme data values. A statistic that is also used to indicate the center of a data set but that is not affected by extreme values is the sample median, deﬁned as the middle value when the data are ranked in order from smallest to largest. We will let m denote the sample median. Deﬁnition Order the data values from smallest to largest. If the number of data values is odd, then the sample median is the middle value in the ordered list; if it is even, then the sample median is the average of the two middle values. It follows from this deﬁnition that if there are three data values, then the sample median is the second-smallest value; and if there are four, then it is the average of the second- and the third-smallest values. ■

Example 3.6 The following data represent the number of weeks it took seven individuals to obtain their driver’s licenses. Find the sample median. 2, 110, 5, 7, 6, 7, 3

83

84

CH A P T E R 3: Using Statistics to Summarize Data Sets

Solution First arrange the data in increasing order. 2, 3, 5, 6, 7, 7, 110 Since the sample size is 7, it follows that the sample median is the fourthsmallest value. That is, the sample median number of weeks it took to obtain a driver’s license is m = 6 weeks. ■ ■

Example 3.7 The following data represent the number of days it took 6 individuals to quit smoking after completing a course designed for this purpose. 1, 2, 3, 5, 8, 100 What is the sample median? Solution Since the sample size is 6, the sample median is the average of the two middle values; thus, m=

3+5 =4 2

That is, the sample median is 4 days.

■

In general, for a data set of n values, the sample median is the [(n + 1)/2]smallest value when n is odd and is the average of the (n/2)-smallest value and the (n/2 + 1)-smallest value when n is even. The sample mean and sample median are both useful statistics for describing the central tendency of a data set. The sample mean, being the arithmetic average, makes use of all the data values. The sample median, which makes use of only one or two middle values, is not affected by extreme values. ■

Example 3.8 The following data give the names of the National Basketball Association (NBA) individual scoring champions and their season scoring averages in each of the seasons from 1992 to 2008. (a) Find the sample median of the scoring averages. (b) Find the sample mean of the scoring averages.

3.3 Sample Median

1992–93

Michael Jordan, Chicago Bulls

32.6

1993–94

David Robinson, San Antonio Spurs

29.8

1994–95

Shaquille O’Neal, Orlando Magic

29.3

1995–96

Michael Jordan, Chicago Bulls

30.4

1996–97

Michael Jordan, Chicago Bulls

29.6

1997–98

Michael Jordan, Chicago Bulls

28.7

1998–99

Allen Iverson, Philadelphia 76ers

26.8

1999–00

Shaquille O’Neal, L.A. Lakers

29.7

2000–01

Allen Iverson, Philadelphia 76ers

31.1

2001–02

Allen Iverson, Philadelphia 76ers

31.4

2002–03

Tracy McGrady, Orlando Magic

32.1

2003–04

Tracy McGrady, Orlando Magic

28.0

2004–05

Allen Iverson, Philadelphia 76ers

30.7

2005–06

Kobe Bryant, L.A. Lakers

35.4

2006–07

Kobe Bryant, Los Angeles Lakers

31.6

2007–08

Lebron James, Cleveland Cavaliers

30.0

2008–09

Dwyane Wade, Miami Heat

30.2

Solution (a) Since there are 17 data values, the sample median is the 9th smallest. Therefore, the sample median is m = 30.2 (b) The sum of all 17 values is 517.4, and so the sample mean is x=

517.4 ≈ 30.435 17

■

Historical Perspective The Dutch mathematician Christian Huyghens was one of the early developers of the theory of probability. In 1669 his brother Ludwig, after studying the mortality tables of the time, wrote to his famous older brother that “I have just been making a table showing how long people have to live. . . . Live well! According to my calculations you will live to be about 56 12 and I to 55.” Christian, intrigued, also looked at the mortality tables but came up with different estimates for how long both he and his brother would live. Why? Because they were looking at different statistics. Ludwig was basing his estimates on the sample median while Christian was basing his on the sample mean!

85

86

CH A P T E R 3: Using Statistics to Summarize Data Sets

For data sets that are roughly symmetric about their central values, the sample mean and sample median will have values close to each other. For instance, the data 4, 6, 8, 8, 9, 12, 15, 17, 19, 20, 22 are roughly symmetric about the value 12, which is the sample median. The sample mean is x = 140/11 = 12.73, which is close to 12. The question as to which of the two summarizing statistics is the more informative depends on what you are interested in learning from the data set. For instance, if a city government has a ﬂat-rate income tax and is trying to ﬁgure out how much income it can expect, then it would be more interested in the sample mean of the income of its citizens than in the sample median (why is this?). On the other hand, if the city government were planning to construct some middle-income housing and were interested in the proportion of its citizens who would be able to afford such housing, then the sample median might be more informative (why is this?). Although it is interesting to consider whether the sample mean or sample median is more informative in a particular situation, note that we need never restrict ourselves to a knowledge of just one of these quantities. They are both important, and thus both should always be computed when a data set is summarized.

PROBLEMS 1. The following are the total yardages of a sample of 12 municipal golf courses: 7040, 6620, 6050, 6300, 7170, 5990, 6330, 6780, 6540, 6690, 6200, 6830 (a) Find the sample median. (b) Find the sample mean. 2. (a) Determine the sample median of the data set 14, 22, 8, 19, 15, 7, 8, 13, 20, 22, 24, 25, 11, 9, 14 (b) Increase each value in (a) by 5, and ﬁnd the new sample median. (c) Multiply each value in (a) by 3, and ﬁnd the new sample median. 3. If the median of the data set xi , i = 1, . . . , n, is 10, what is the median of the data set 2xi + 3, i = 1, . . . , n? 4. The following are the speeds of 40 cars as measured by a radar device on a city street:

3.3 Sample Median

22, 26, 31, 38, 27, 29, 33, 40, 36, 27, 25, 42, 28, 19, 28, 26, 33, 26, 37, 22, 31, 30, 44, 29, 25, 17, 46, 28, 31, 29, 40, 38, 26, 43, 45, 21, 29, 36, 33, 30 Find the sample median. 5. The following presents the male and female suicide rates per 100,000 population for a variety of countries. Suicide Rates per 100,000 Population Sex

United States

Australia

Austria

Canada

Denmark

France

5.4 19.7

5.1 18.2

15.8 42.1

5.4 20.5

20.6 35.1

12.7 33.1

Female Male

Sex

Italy Japan Netherlands Poland Sweden U.K. W. Germany

Female 4.3 Male 11.0

14.9 27.8

8.1 14.6

4.4 22.0

11.5 25.0

5.7 12.1

12.0 26.6

Source: World Health Organization, World Health Statistics.

(a) (b) (c) (d)

Find the sample median of the male suicide rates. Find the sample median of the female suicide rates. Find the sample mean of the male suicide rates. Find the sample mean of the female suicide rates.

6. Find the sample median of the average annual number of days of precipitation in the cities noted in Prob. 3 of Sec. 3.2. 7. Find the sample median of the average annual number of inches of precipitation in the cities noted in Prob. 3 of Sec. 3.2. 8. Find the sample median of the data presented in Prob. 8 of Sec. 2.3. 9. Use the table on death rates preceding Prob. 9 of Sec. 2.3 to ﬁnd the sample median of the death rates due to (a) Falls (b) Poisoning (c) Drowning 10. The sample median of 10 distinct values is 5. What can you say about the new sample median if (a) An additional datum whose value is 7 is added to the data set? (b) Two additional data values—3 and 42—are added to the data set? 11. The histogram in the ﬁgure on the following page describes the annual rainfall, in inches, over the last 34 years in a certain western city. Since the raw data are not recoverable from a histogram, we cannot use them to exactly compute the value of the sample mean and sample median. Still, based on this histogram, what is the largest possible value of

87

88

CH A P T E R 3: Using Statistics to Summarize Data Sets

(a) The sample mean? (b) The sample median? What is the smallest possible value of (c) The sample mean? (d) The sample median? (e) The actual data follow: 15.2, 16.1, 16.5, 16.7, 17.2, 17.5, 17.7, 18.3, 18.6, 18.8, 18.9, 19.1, 19.2, 19.2, 19.6, 19.8, 19.9, 20.2, 20.3, 20.3, 20.8, 21.1, 21.4, 21.7, 22.2, 22.5, 22.5, 22.7, 22.9, 23.3, 23.6, 24.1, 24.5, 24.9 Determine the sample mean and sample median and see that they are consistent with your previous answers. 12. A total of 100 people work at company A, whereas a total of 110 work at company B. Suppose the total employee payroll is larger at company A than at company B. (a) What does this imply about the sample mean of the salaries at company A with regards to the sample mean of the salaries at company B? (b) What does this imply about the sample median of the salaries at company A with regards to the sample median of the salaries at company B? 13. Using the data from Example 3.4, compute the sample medians of the severity of head injuries suffered by motorcycle operators who were wearing and who were not wearing helmets.

3.3 Sample Median

14. In the following situations, which do you think is a more informative statistic, the sample mean or the sample median? (a) In order to decide whether to discontinue a bus service from Rochester to New York City, an executive studies the number of riders on a sample of days. (b) To determine how present-day college-bound students compare with those of earlier years, a sample of entrance examination scores from several years is consulted. (c) A lawyer representing a defendant in a jury trial is studying the IQ scores of the jurors who were selected. (d) You purchased your home 6 years ago in a small suburban community for $105,000, which was both the mean and the median price for all homes sold that year in that community. However, in the last couple of years some new, more expensive homes have been built. To get an idea of the present value of your home, you study recent sales prices of homes in your community. 15. Women make up the following percentages of the workforce in the 14 occupations listed.

Occupation Corporate executives Nurses Sales supervisors Sales workers Firefighters Cleaning jobs Construction workers

Percentage women 36.8 94.3 30.5 68.6 1.9 41.5 2.8

Occupation Doctors Lawyers Elementary school teachers Postal clerks Police workers Construction supervisors Truck drivers

Percentage women 17.6 18.0 85.2 43.5 10.9 1.6 2.1

For these percentages ﬁnd (a) The sample mean (b) The sample median It also turns out that women make up 44.4 percent of the total workforce for these occupations. Is this consistent with your answers in (a) and (b)? Explain! 16. Using data concerning the ﬁrst 30 students in App. A, ﬁnd the sample median and the sample mean for (a) Weight (b) Cholesterol (c) Blood pressure 17. The following table gives the median age at ﬁrst marriage in the years 1992 to 2002.

89

90

CH A P T E R 3: Using Statistics to Summarize Data Sets

(a) Find the sample median of the men’s median age. (b) Find the sample median of the women’s median age. U.S. Median Age at First Marriage Year

Men

Women

Year

Men

Women

2002

26.9

25.3

1996

27.1

24.8

2001

26.9

25.1

1995

26.9

24.5

2000

26.8

25.1

1994

26.7

24.5

1999

26.9

25.1

1993

26.5

24.5

1998

26.7

25.0

1992

26.5

24.4

1997

26.8

25.0

3.3.1 Sample Percentiles The sample median is a special type of statistic known as a sample 100p percentile, where p is any fraction between 0 and 1. Loosely speaking, a sample 100p percentile is the value such that 100p percent of the data values are less than it and 100(1 − p) percent of the values are greater than it. Deﬁnition The sample 100p percentile is that data value having the property that at least 100p percent of the data are less than or equal to it and at least 100(1 − p) percent of the data values are greater than or equal to it. If two data values satisfy this condition, then the sample 100p percentile is the arithmetic average of these values. Note that the sample median is the sample 50th percentile. That is, it is the sample 100p percentile when p = 0.50. Suppose the data from a sample of size n are arranged in increasing order from smallest to largest. To determine the sample 100p percentile, we must determine the data value such that 1. At least np of the data values are less than or equal to it. 2. At least n(1 − p) of the data values are greater than or equal to it. Now if np is not an integer, then the only data value satisfying these requirements is the one whose position is the smallest integer greater than np. For instance, suppose we want the sample 90th percentile from a sample of size n = 12. Since p = 0.9, we have np = 10.8 and n(1 − p) = 1.2. Thus, we require those data values for which 1. At least 10.8 values are less than or equal to it (and so the data value must be in position 11 or higher). 2. At least 1.2 values are greater than or equal to it (and so it must be in position 11 or lower).

3.3 Sample Median

Clearly, the only data value that satisﬁes both requirements is the one that is in position 11, and thus this is the sample 90th percentile. On the other hand, if np is an integer, then both the value in position np and the value in position np + 1 satisfy requirements (1) and (2); and so the sample 100p percentile value would be the average of these two data values. For instance, suppose we wanted the sample 95th percentile from a data set of n = 20 values. Then both the 19th and the 20th values (that is, the two largest values) will be greater than or equal to at least np = 20(0.95) = 19 of the values and less than or equal to at least n(1 − p) = 1 value. The 95th percentile is thus the average of the 19th and 20th largest values. Summing up, we have shown the following. To ﬁnd the sample 100p percentile of a data set of size n 1. Arrange the data in increasing order. 2. If np is not an integer, determine the smallest integer greater than np. The data value in that position is the sample 100p percentile. 3. If np is an integer, then the average of the values in positions np and np + 1 is the sample 100p percentile.

■

Example 3.9 Which data value is the sample 90th percentile when the sample size is (a) 8, (b) 16, and (c) 100? Solution (a) Since 0.9 × 8 = 7.2, which is not an integer, it follows that if the data are arranged from smallest to largest, then the sample 90th percentile value would be the 8th-smallest value (that is, the largest value). (b) Since 0.9 × 16 = 14.4, which is not an integer, it follows that the sample 90th percentile would be the 15th-smallest value. (c) Since 0.9 × 100 = 90 is an integer, the sample 90th percentile value is the average of the 90th and the 91st values when the data are arranged from smallest to largest. ■

■

Example 3.10 Table 3.1 lists the top 20 U.S. colleges and universities based on endowment assets. Using these data, ﬁnd the (a) Sample 90th percentile (b) Sample 20th percentile

91

92

CH A P T E R 3: Using Statistics to Summarize Data Sets

Table 3.1 Top 20 Colleges and Universities in Endowment Assets, 2005

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Institution

State

2005 Endowment Funds ($000)

Harvard University Yale University Stanford University University of Texas System Princeton University Massachusetts Institute of Technology University of California Columbia University The Texas A&M University System and Foundations University of Michigan Emory University University of Pennsylvania Washington University Northwestern University University of Chicago Duke University Cornell University University of Notre Dame Rice University University of Virginia

MA CT CA TX NJ MA CA NY TX

25,473,721 15,224,900 12,205,000 11,610,997 11,206,500 6,712,436 5,221,916 5,190,564 4,963,879

Ml GA PA MO IL IL NC NY IN TX VA

4,931,338 4,376,272 4,369,782 4,268,415 4,215,275 4,137,494 3,826,153 3,777,092 3,650,224 3,611,127 3,219,098

Solution (a) Because the sample size is 20 and 20 × 0.9 = 18, the sample 90th percentile is the average of the 18th- and 19th-smallest values. Equivalently, it is the average of the 2nd- and 3rd-largest values. Hence, sample 90th percentile =

15,224,900 + 12,205,000 = 13,714,950 2

That is, the sample 90th percentile of this data set is approximately $13.7 billion. (b) Because 20 × 0.2 = 4, the sample 20th percentile is the average of the 4th- and 5th-smallest values, giving the result Sample 20th percentile =

3,777,092 + 3,826,153 = 3,801,623 2

■

3.3 Sample Median

The sample 25th percentile, 50th percentile, and 75th percentile are known as the quartiles. Deﬁnition The sample 25th percentile is called the ﬁrst quartile. The sample 50th percentile is called the median or the second quartile. The sample 75th percentile is called the third quartile. The quartiles break up a data set into four parts with about 25 percent of the data values being less than the ﬁrst quartile, about 25 percent being between the ﬁrst and second quartiles, about 25 percent being between the second and third quartiles, and about 25 percent being larger than the third quartile.

■

Example 3.11 Find the sample quartiles for the following 18 data values, which represent the ordered values of a sample of scores from a league bowling tournament: 122, 126, 133, 140, 145, 145, 149, 150, 157, 162, 166, 175, 177, 177, 183, 188, 199, 212 Solution Since 0.253 × 18 = 4.5, the sample 25th percentile is the ﬁfth-smallest value, which is 145. Since 0.50 × 18 = 9, the second quartile (or sample median) is the average of the 9th- and 10th-smallest values and so is 157 + 162 = 159.5 2 Since 0.75 × 18 = 13.5, the third quartile is the 14th-smallest value, which is 177. ■

PROBLEMS 1. Seventy-ﬁve values are arranged in increasing order. How would you determine the sample (a) 80th percentile (b) 60th percentile (c) 30th percentile of this data set? 2. The following table gives the number of deaths of infants per 1,000 births in the 50 U.S. states in 2007. Use it to ﬁnd the quartiles of the state infant death rates.

93

94

CH A P T E R 3: Using Statistics to Summarize Data Sets

Number of Deaths of Infants per 1,000 Births and Total Infant Mortality State Ranking 1 Montana 1 Vermont 3 Minnesota 4 Massachusetts 5 Iowa 6 California 6 Utah 8 Rhode Island 9 Connecticut 9 Oregon 9 Washington 12 New Hampshire 12 New Jersey 12 North Dakota 15 Hawaii 15 Maine 17 Wisconsin 18 New York 19 Idaho 20 Colorado 20 New Mexico 20 Texas 23 Nevada 24 Nebraska 25 Alaska

4.5 (52 Total) 4.5 (30 Total) 4.7 (332 Total) 4.8 (380 Total) 5.1 (195 Total) 5.2 (2,811 Total) 5.2 (264 Total) 5.3 (68 Total) 5.5 (233 Total) 5.5 (251 Total) 5.5 (451 Total) 5.6 (81 Total) 5.6 (651 Total) 5.6 (46 Total) 5.7 (104 Total) 5.7 (79 Total) 6.0 (420 Total) 6.1 (1,518 Total) 6.2 (139 Total) 6.3 (434 Total) 6.3 (179 Total) 6.3 (2,407 Total) 6.4 (225 Total) 6.6 (173 Total) 6.7 (69 Total)

25 Arizona 27 Kentucky 28 Florida 29 Kansas 29 Pennsylvania 31 Illinois 31 Missouri 31 Virginia 34 Michigan 34 West Virginia 36 Ohio 37 Indiana 37 Oklahoma 39 South Dakota 40 Arkansas 41 Maryland 42 Georgia 43 Delaware 43 Tennessee 45 Alabama 46 North Carolina 46 Wyoming 48 South Carolina 49 Mississippi 50 Louisiana

6.7 (630 Total) 6.8 (378 Total) 7.0 (1,537 Total) 7.2 (284 Total) 7.2 (1,049 Total) 7.5 (1,349 Total) 7.5 (584 Total) 7.5 (776 Total) 7.6 (984 Total) 7.6 (158 Total) 7.7 (1,143 Total) 8.0 (700 Total) 8.0 (411 Total) 8.2 (93 Total) 8.3 (319 Total) 8.4 (630 Total) 8.5 (1,181 Total) 8.6 (98 Total) 8.6 (687 Total) 8.7 (516 Total) 8.8 (1,053 Total) 8.8 (60 Total) 9.3 (525 Total) 9.8 (420 Total) 10.5 (684 Total)

3. Consider a data set of n values 1, 2, 3, . . . , n. Find the value of the sample 95th percentile when (a) n = 100 (b) n = 101 The following table gives the number of physicians and of dentists per 100,000 population for 12 midwestern states in 2000. Problems 4 and 5 are based on it. 4. For the physician’s rates per 100,000 population, ﬁnd the (a) Sample 40th percentile (b) Sample 60th percentile (c) Sample 80th percentile

3.3 Sample Median

State Ohio Indiana Illinois Michigan Wisconsin Minnesota Iowa Missouri North Dakota South Dakota Nebraska Kansas

Physician’s rate

Dentist’s rate

188 146 206 177 177 207 141 186 157 129 162 166

56 48 61 64 70 70 60 55 55 54 71 52

Source: American Medical Association, Physician Characteristics and Distribution in the U.S.

5. For the dentist’s rates per 100,000 population, ﬁnd the (a) Sample 90th percentile (b) Sample 50th percentile (c) Sample 10th percentile 6. Suppose the sample 100p percentile of a set of data is 120. If we add 30 to each data value, what is the new value of the sample 100p percentile? 7. Suppose the sample 100p percentile of a set of data is 230. If we multiply each data value by a positive constant c, what is the new value of the sample 100p percentile? 8. Find the sample 90th percentile of this data set: 75, 33, 55, 21, 46, 98, 103, 88, 35, 22, 29, 73, 37, 101, 121, 144, 133, 52, 54, 63, 21, 7 9. Use the table on page 96 to ﬁnd the quartiles of 2006 trafﬁc fatality rates (per 100 million vehicle miles) in the 50 states of the United States. 10. The following are the quartiles of a large data set: First quartile = 35 Second quartile = 47 Third quartile = 66

95

Trafﬁc Fatalities by State: 1990 to 2006 [For deaths within 30 days of the accident] Fatality rate1 State

1990

2000

2005

2006

1990

2006

U.S… AL …. AK…. AZ…. AR …. CA….

44,599 1,121 98 869 604 5,192

41,945 996 106 1,036 652 3,753

43,510 1,148 73 1,179 654 4,333

42,642 1,208 74 1,288 665 4,236

2.1 2.6 2.5 2.5 2.9 2.0

CO…. CT…. DE …. DC…. FL…..

544 385 138 48 2,891

681 341 123 48 2,999

606 278 133 48 3,518

535 301 148 37 3,374

GA…. HI….. ID….. IL….. IN…..

1,562 177 244 1,589 1,049

1,541 132 276 1,418 886

1,729 140 275 1,363 938

IA….. KS…. KY…. LA…. ME….

465 444 849 959 213

445 461 820 938 169

MD…. MA…. Ml….. MN…. MS….

707 605 1,571 566 750

588 433 1,382 625 949

1 Deaths

Fatality rate1 State

1990

2000

2005

2006

1990

2006

1.4 2.0 1.5 2.1 2.0 1.3

MO…. MT…. NE …. NV …. NH …. NJ…..

1,097 212 262 343 158 886

1,157 237 276 323 126 731

1,257 251 276 427 166 747

1,096 263 269 432 127 772

2.2 2.5 1.9 3.4 1.6 1.5

1.6 2.3 1.4 2.0 0.9 1.0

2.0 1.5 2.1 1.4 2.6

1.1 1.0 1.6 1.0 1.7

NM…. NY…. NC…. ND…. OH….

499 2,217 1,385 112 1,638

432 1,460 1,557 86 1,366

488 1,434 1,547 123 1,321

484 1,456 1,559 111 1,238

3.1 2.1 2.2 1.9 1.8

1.9 1.0 1.5 1.4 1.1

1,693 161 267 1,254 899

2.2 2.2 2.5 1.9 2.0

1.5 1.6 1.8 1.2 1.3

OK…. OR…. PA…. RI….. SC….

641 579 1,646 84 979

650 451 1,520 80 1,065

803 487 1,616 87 1,094

765 477 1,525 81 1,037

1.9 2.2 1.9 1.1 2.8

1.6 1.3 1.4 1.0 2.1

450 428 985 963 169

439 468 913 982 188

2.0 1.9 2.5 2.5 1.8

1.4 1.6 1.9 2.2 1.3

SD…. TN…. TX…. UT …. VT….

153 1,177 3,250 272 90

173 1,307 3,779 373 76

186 1,270 3,536 282 73

191 1,287 3,475 287 87

2.2 2.5 2.1 1.9 1.5

2.1 1.8 1.5 1.1 1.1

614 441 1,129 559 931

651 430 1,085 494 911

1.7 1.3 1.9 1.5 3.1

1.2 0.8 1.0 0.9 2.2

VA…. WA…. WV…. Wl….. WY….

1,079 825 481 769 125

929 631 411 799 152

947 649 374 815 170

963 630 410 724 195

1.8 1.8 3.1 1.7 2.1

1.2 1.1 2.0 1.2 2.1

per 100 million vehicle miles traveled. Source: U.S. National Highway Traffic Safety Administration, Trafﬁc Safety Facts, annual. See .

3.4 Sample Mode

(a) Give an interval in which approximately 50 percent of the data lie. (b) Give a value which is greater than approximately 50 percent of the data. (c) Give a value such that approximately 25 percent of the data values are greater than it. 11. A symmetric data set has its median equal to 40 and its third quartile equal to 55. What is the value of the ﬁrst quartile?

3.4 SAMPLE MODE Another indicator of central tendency is the sample mode, which is the data value that occurs most frequently in the data set. ■

Example 3.12 The following are the sizes of the last 8 dresses sold at a women’s boutique: 8, 10, 6, 4, 10, 12, 14, 10 What is the sample mode? Solution The sample mode is 10, since the value of 10 occurs most frequently.

■

If no single value occurs most frequently, then all the values that occur at the highest frequency are called modal values. In such a situation we say that there is no unique value of the sample mode. ■

Example 3.13 The ages of 6 children at a day care center are 2, 5, 3, 5, 2, 4 What are the modal values of this data set? Solution Since the ages 2 and 5 both occur most frequently, both 2 and 5 are modal values. ■

■

Example 3.14 The following frequency table gives the values obtained in 30 throws of a die.

97

98

CH A P T E R 3: Using Statistics to Summarize Data Sets

Value 1 2 3 4 5 6

Frequency 6 4 5 8 3 4

It is easy to pick out the modal value from a frequency table, since it is just that value having the largest frequency. For these data, ﬁnd the (a) Sample mode (b) Sample median (c) Sample mean Solution (a) Since the value 4 appears with the highest frequency, the sample mode is 4. (b) Since there are 30 data values, the sample median is the average of the 15th- and 16th-smallest values. Since the 15th-smallest value is 3 and the 16th-smallest is 4, the sample median is 3.5. (c) The sample mean is x=

1·6+2·4+3·5+4·8+5·3+6·4 100 = ≈ 3.333 30 30

■

PROBLEMS 1. Match each statement in the left-hand column with the correct data set from the right-hand column. 1. Sample mode is 9 2. Sample mean is 9 3. Sample median is 9

A: 5, 7, 8, 10, 13, 14 B: 1, 2, 5, 9, 9, 15 C: 1, 2, 9, 12, 12, 18

2. Using the data from Example 2.2, ﬁnd the sample mode of the winning Masters Golf Tournament scores. 3. Using data concerning the ﬁrst 100 students in App. A, ﬁnd the sample mode for (a) Weight (b) Blood pressure (c) Cholesterol

3.5 Sample Variance and Sample Standard Deviation

4. Suppose you want to guess the salary of a bank vice president whom you have just met. If you want to have the greatest chance of being correct to the nearest $1000, would you rather know the sample mean, the sample median, or the sample mode of the salaries of bank vice presidents? 5. Construct a data set for which the sample mean is 10, the sample median is 8, and the sample mode is 6. 6. If the sample mode of the data xi , i = 1, . . . , n, is equal to 10, what is the sample mode of the data yi = 2xi + 5, i = 1, . . . , n? 7. Joggers use a quarter-mile track around an athletic ﬁeld. In a sample of 17 joggers, 1 did 2 loops, 4 did 4 loops, 5 did 6 loops, 6 did 8 loops, and 1 did 12 loops. (a) What is the sample mode of the number of loops run by these joggers? (b) What is the sample mode of the distances run by these joggers? 8. The sample mean, sample median, and sample mode of the ﬁrst 99 values of a data set of 198 values are all equal to 120. If the sample mean, median, and mode of the ﬁnal 99 values are all equal to 100, what can you say about the sample mean of the entire data set? What can you say about the sample median? What about the sample mode?

3.5 SAMPLE VARIANCE AND SAMPLE STANDARD DEVIATION Whereas so far we have talked about statistics that measure the central tendency of a data set, we have not yet considered ones that measure its spread or variability. For instance, although the following data sets A and B have the same sample mean and sample median, there is clearly more spread in the values of B than in those of A. A: 1, 2, 5, 6, 6

B: −40, 0, 5, 20, 35

One way of measuring the variability of a data set is to consider the deviations of the data values from a central value. The most commonly used central value for this purpose is the sample mean. If the data values are x1 , . . . , xn and the sample mean is x = ni=1 xi /n then the deviation of the value xi from the sample mean is xi − x, i = 1, . . . , n. One might suppose that a natural measure of the variability of a set of data would be the average of the deviations from the mean. However, as we have shown in Sec. 3.2, ni=1 (xi − x) = 0. That is, the sum of the deviations from the sample mean is always equal to 0, and thus the average of the deviations from the sample mean must also be 0. However, after some additional reﬂection it should be clear that we really do not want to allow the positive and the negative deviations to

99

100

CH A P T E R 3: Using Statistics to Summarize Data Sets

cancel. Instead, we should be concerned about the individual deviations without regard to their signs. This can be accomplished either by considering the absolute values of the deviations or, as turns out to be more useful, by considering their squares. The sample variance is a measure of the “average” of the squared deviations from the sample mean. However, for technical reasons (which will become clear in Chap. 8) this “average” divides the sum of the n squared deviations by the quantity n − 1, rather than by the usual value n. Deﬁnition sample variance, call it s2 , of the data set xi , . . . , xn having sample The n mean x = i=1 xi n is deﬁned by n (xi − x)2 2 s = i=1 n−1 ■

Example 3.15 Find the sample variance of data set A. Solution It is determined as follows: xi x xi − x (xi − x)2

1 4 −3 9

2 4 −2 4

5 4 1 1

6 4 2 4

6 4 2 4

Hence, for data set A, s2 =

■

9+4+1+4+4 = 5.5 4

■

Example 3.16 Find the sample variance for data set B. Solution The sample mean for data set B is also x = 4. Therefore, for this set, we have xi xi − x (xi − x)2

−40 −44 1936

0 −4 16

5 1 1

20 16 256

35 31 961

3.5 Sample Variance and Sample Standard Deviation

Thus, s2 =

3170 = 792.5 4

■

The following algebraic identity is useful for computing the sample variance by hand: n

(xi − x)2 =

i=1

■

n

xi2 − nx 2

(3.2)

i=1

Example 3.17 Check that identity (3.2) holds for data set A. Solution Since n = 5 and x = 4, 5

xi2 − nx 2 = 1 + 4 + 25 + 36 + 36 − 5(16) = 102 − 80 = 22

i=1

From Example 3.15, 5

(xi − x)2 = 9 + 4 + 1 + 4 + 4 = 22

i=1

■

and so the identity checks out.

Suppose that we add a constant c to each of the data values x1 , . . . , xn to obtain the new data set y1 , . . . , yn , where yi = xi + c To see how this affects the value of the sample variance, recall from Sec. 3.2 that y =x+c and so yi − y = xi + c − (x + c) = xi − x That is, the y deviations are equal to the x deviations, and therefore their sums of squares are equal. Thus, we have shown the following useful result.

101

102

CH A P T E R 3: Using Statistics to Summarize Data Sets

The sample variance remains unchanged when a constant is added to each data value. The preceding result can often be used in conjunction with the algebraic identity (3.2) to greatly reduce the time it takes to compute the sample variance.

■

Example 3.18 The following data give the yearly numbers of law enforcement ofﬁcers killed in the United States over 10 years: 164, 165, 157, 164, 152, 147, 148, 131, 147, 155 Find the sample variance of the number killed in these years. Solution Rather than working directly with the given data, let us subtract the value 150 from each data item. (That is, we are adding c = −150 to each data value.) This results in the new data set 14, 15, 7, 14, 2, −3, −2, −19, −3, 5 Its sample mean is y=

14 + 15 + 7 + 14 + 2 − 3 − 2 − 19 − 3 + 5 = 3.0 10

The sum of the squares of the new data is 10

yi2 = 142 + 152 + 72 + 142 + 22 + 32 + 22 + 192 + 32 + 52 = 1078

i=1

Therefore, using the algebraic identity (3.2) shows that 10

(yi − y)2 = 1078 − 10(9) = 988

i=1

Hence, the sample variance of the revised data, which is equal to the sample variance of the original data, is s2 =

988 ≈ 109.78 9

■

The positive square root of the sample variance is called the sample standard deviation.

3.5 Sample Variance and Sample Standard Deviation

Deﬁnition The quantity s, deﬁned by s=

n i=1 (xi

− x)2 n−1

is called the sample standard deviation. The sample standard deviation is measured in the same units as the original data. That is, for instance, if the data are in feet, then the sample variance will be expressed in units of square feet and the sample standard deviation in units of feet. If each data value xi , i = 1, . . . , n, is multiplied by a constant c to obtain the new data set yi = cxi

i = 1, . . . , n

then the sample variance of the y data is the sample variance of the x data multiplied by c 2 . That is, sy2 = c 2 sx2 where sy2 and sx2 are the sample variances of the new and old data sets, respectively. Taking the square root of both sides of the preceding equation shows that the standard deviation of the y data is equal to the absolute value of c times the standard deviation of the x data, or sy = |c|sx Another indicator of the variability of a data set is the interquartile range, which is equal to the third minus the ﬁrst quartile. That is, roughly speaking, the interquartile range is the length of the interval in which the middle half of the data values lie. ■

Example 3.19 The Miller Analogies Test is a standardized test that is taken by a variety of students applying to graduate and professional schools. Table 3.2 presents some of the percentile scores on this examination for students, classiﬁed according to the graduate ﬁelds they are entering. For instance, Table 3.2 states that the median grade of students in the physical sciences is 68, whereas it is 49 for those applying to law school. Determine the interquartile ranges of the scores of students in the ﬁve speciﬁed categories.

103

104

CH A P T E R 3: Using Statistics to Summarize Data Sets

Table 3.2 Selected Percentiles on the Miller Analogies Test for Five Categories of Students Percentile 99 90 75 50 25

Physical sciences

Medical school

Social sciences

Languages and literature

Law school

93 88 80 68 55

92 78 71 57 45

90 82 74 61 49

87 80 73 59 43

84 73 60 49 37

Solution Since the interquartile range is the difference between the 75th and the 25th sample percentiles, it follows that its value is 80 − 55 = 25 71 − 45 = 26 74 − 49 = 25 73 − 43 = 30 60 − 37 = 23

for scores of physical science students for scores of medical school students for scores of social science students for scores of language and literature students for scores of law school students

■

A box plot is often used to plot some of the summarizing statistics of a data set. A straight-line segment stretching from the smallest to the largest data value is drawn on a horizontal axis; imposed on the line is a “box,” which starts at the ﬁrst and continues to the third quartile, with the value of the second quartile indicated by a vertical line. For instance, the following frequency table gives the starting salaries of a sample of 42 graduating seniors of a liberal arts college. Starting salary 47 48 49 50 51 52 53 54 56 57 60

Frequency 4 1 3 5 8 10 0 5 2 3 1

The salaries go from a low of 47 to a high of 60. The value of the ﬁrst quartile (equal to the value of the 11th smallest on the list) is 50; the value of the second

3.5 Sample Variance and Sample Standard Deviation

quartile (equal to the average of the 21st- and 22nd-smallest values) is 51.5; and the value of the third quartile (equal to the value of the 32nd smallest on the list) is 54. The box plot for this data set is as follows.

PROBLEMS 1. The following data give the per capita consumption of milk in the years from 1983 to 1987. The data are from the U.S. Department of Agriculture, Food Consumption, Prices, and Expenditures, annual. Year

Amount (in gallons per capita)

1983

26.3

1984

26.2

1985

26.4

1986

26.3

1987

25.9

Find the sample mean and the sample variance of this set. 2. You are given these data sets: A: 66, 68, 71, 72, 72, 75

B: 2, 5, 9, 10, 10, 16

(a) Which one appears to have the larger sample variance? (b) Determine the sample variance of data set A. (c) Determine the sample variance of data set B. 3. The Masters Golf Tournament and the U.S. Open are the two most prestigious golf tournaments in the United States. The Masters is always played on the Augusta National golf course, whereas the U.S. Open is played on different courses in different years. As a result, one might expect the sample variance of the winning scores in the U.S. Open to be higher than that of the winning scores in the Masters. To check whether this is so, we have collected the winning scores in both tournaments for 1981 to 1990.

105

106

CH A P T E R 3: Using Statistics to Summarize Data Sets

Winning score Tournament

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

U.S. Open Masters

273 280

282 284

280 280

276 277

279 282

279 279

277 285

278 281

278 283

280 278

(a) Compute the sample variance of the winning scores in the U.S. Open tournament. (b) Compute the sample variance of the winning scores in the Masters tournament. The following table gives the numbers of physicians and dentists in Japan in the even-numbered years between 1984 and 2000. Problems 4 and 5 are based on this table. Number of Physicians and Dentists (1984–2000)

1984 1986 1988 1990 1992 1994 1996 1998 2000

Physicians

Dentists

173,452 183,129 193,682 203,797 211,498 220,853 230,297 236,933 243,201

61,283 64,904 68,692 72,087 75,628 79,091 83,403 85,669 88,410

4. Just by eyeballing, estimate the ratio of the sample variance of the yearly number of physicians to the sample variance of the yearly number of dentists. 5. Find the actual value of the ratio in problem 4. 6. An individual needing automobile insurance requested quotes from 10 different insurers for identical coverage and received the following values (amounts are annual premiums in dollars): 720, 880, 630, 590, 1140, 908, 677, 720, 1260, 800 Find (a) The sample mean (b) The sample median (c) The sample standard deviation The following table gives the 2008 populations for each U.S. state and territory. Problems 7, 8, and 9 refer to this data.

3.5 Sample Variance and Sample Standard Deviation

7. 8. 9. 10.

State

Population (2008)

State

California Texas New York Florida Illinois Pennsylvania Ohio Michigan Georgia North Carolina New Jersey Virginia Washington Arizona Massachusetts Indiana Tennessee Missouri Maryland Wisconsin Minnesota Colorado Alabama South Carolina Louisiana Kentucky Puerto Rico Oregon

36,756,666 24,326,974 19,490,297 18,328,340 12,901,563 12,448,279 11,485,910 10,003,422 9,685,744 9,222,414 8,682,661 7,769,089 6,549,224 6,500,180 6,497,967 6,376,792 6,214,888 5,911,605 5,633,597 5,627,967 5,220,393 4,939,456 4,661,900 4,479,800 4,410,796 4,269,245 3,954,037 3,790,060

Oklahoma Connecticut Iowa Mississippi Arkansas Kansas Utah Nevada New Mexico West Virginia Nebraska Idaho Maine New Hampshire Hawaii Rhode Island Montana Delaware South Dakota Alaska North Dakota Vermont District of Columbia Wyoming Guam US Virgin Islands Northern Mariana Islands American Samoa

Population (2008) 3,642,361 3,501,252 3,002,555 2,938,618 2,855,390 2,802,134 2,736,424 2,600,167 1,984,356 1,814,468 1,783,432 1,523,816 1,316,456 1,315,809 1,288,198 1,050,788 967,440 873,092 804,194 686,293 641,481 621,270 591,833 532,668 173,456 108,448 84,546 57,291

Find the sample variance of the populations of the ﬁrst 17 locales. Find the sample variance of the populations of the next 17 locales. Find the sample variance of the populations of the ﬁnal 17 locales. If s2 is the sample variance of the data xi , i = 1, . . . , n, what is the sample variance of the data axi + b, i = 1, . . . , n, when a and b are given constants? 11. Compute the sample variance and sample standard deviation of the following data sets: (a) 1, 2, 3, 4, 5 (b) 6, 7, 8, 9, 10 (c) 11, 12, 13, 14, 15 (d) 2, 4, 6, 8, 10 (e) 10, 20, 30, 40, 50

107

108

CH A P T E R 3: Using Statistics to Summarize Data Sets

12. On the U.S. side of the U.S.–Canada border, temperatures are measured in degrees Fahrenheit, whereas on the Canadian side they are measured in degrees Celsius (also called Centigrade). Suppose that during the month of January the sample mean of the temperatures, as recorded on the U.S. side of the border, was 40 ◦ F with a sample variance of 12. Use the formula for converting a Fahrenheit temperature to a Celsius temperature C=

5 (F − 32) 9

to ﬁnd (a) The sample mean recorded by the Canadians (b) The sample variance recorded by the Canadians 13. Compute the sample mean and sample variance of the systolic blood pressures of the ﬁrst 50 students of the data set of App. A. Now do the same with the last 50 students of this data set. Compare your answers. Comment on the results of this comparison. Do you ﬁnd it surprising? 14. If s is the sample standard deviation of the data xi , i = 1, . . . , n, what is the sample standard deviation of axi + b, i = 1, . . . , n? In this problem, a and b are given constants. 15. The following table gives the number of motorcycle retail sales in Japan for 8 different years. Use it to ﬁnd the sample standard deviation of the number of motorcycle sales in the 8 years. Year

2001 2002 2003 2004 2005 2006 2007 2008

Motorcycle sales (in thousands)

751

771

760

700

707

700

685

522

Source: Motorcycle Industry Council.

16. Find the sample standard deviation of the data set given by the following frequency table: Value 3 4

Frequency 1 2

Value 5 6

Frequency 3 2

17. The following data represent the acidity of 40 successive rainfalls in the state of Minnesota. The acidity is measured on a pH scale, which varies from 1 (very acidic) to 7 (neutral).

3.6 Normal Data Sets and the Empirical Rule

3.71, 4.23, 4.16, 2.98, 3.23, 4.67, 3.99, 5.04, 4.55, 3.24, 2.80, 3.44, 3.27, 2.66, 2.95, 4.70, 5.12, 3.77, 3.12, 2.38, 4.57, 3.88, 2.97, 3.70, 2.53, 2.67, 4.12, 4.80, 3.55, 3.86, 2.51, 3.33, 3.85, 2.35, 3.12, 4.39, 5.09, 3.38, 2.73, 3.07 (a) Find the sample standard deviation. (b) Find the range. (c) Find the interquartile range. 18. Consider the following two data sets: A: 4.5, 0, 5.1, 5.0, 10, 5.2

B: 0.4, 0.1, 9, 0, 10, 9.5

(a) Determine the range for each data set. (b) Determine the sample standard deviation for each data set. (c) Determine the interquartile range for each data set.

3.6 NORMAL DATA SETS AND THE EMPIRICAL RULE Many of the large data sets one encounters in practice have histograms that are similar in shape. These histograms are often symmetric about their point of highest frequency and then decrease on both sides of this point in a bell-shaped fashion. Such data sets are said to be normal, and their histograms are called normal histograms. Deﬁnition A data set is said to be normal if a histogram describing it has the following properties: 1. It is highest at the middle interval. 2. Moving from the middle interval in either direction, the height decreases in such a way that the entire histogram is bell-shaped. 3. The histogram is symmetric about its middle interval. Figure 3.2 shows the histogram of a normal data set.

FIGURE 3.2 Histogram of a normal data set.

109

110

CH A P T E R 3: Using Statistics to Summarize Data Sets

FIGURE 3.3 Histogram of an approximately normal data set.

FIGURE 3.4 Histogram of a data set skewed to the left.

If the histogram of a data set is close to being a normal histogram, then we say that the data set is approximately normal. For instance, the histogram given in Fig. 3.3 is from an approximately normal data set, whereas the ones presented in Figs. 3.4 and 3.5 are not (since each is too nonsymmetric). Any data set that is not approximately symmetric about its sample median is said to be skewed. It is called skewed to the right if it has a long tail to the right and skewed to the left if it has a long tail to the left. Thus the data set presented in Fig. 3.4 is skewed to the left, and the one of Fig. 3.5 is skewed to the right. It follows from the symmetry of the normal histogram that a data set that is approximately normal will have its sample mean and sample median approximately equal. Suppose that x and s are the sample mean and sample standard deviation, respectively, of an approximately normal data set. The following rule, known as the empirical rule, speciﬁes the approximate proportions of the data observations that are within s, 2s, and 3s of the sample mean x.

3.6 Normal Data Sets and the Empirical Rule

FIGURE 3.5 Histogram of a data set skewed to the right.

Empirical Rule If a data set is approximately normal with sample mean x and sample standard deviation s, then the following are true. 1. Approximately 68 percent of the observations lie within x±s 2. Approximately 95 percent of the observations lie within x ± 2s 3. Approximately 99.7 percent of the observations lie within x ± 3s

■

Example 3.20 The scores of 25 students on a history examination are listed on the following stem-and-leaf plot. 9 8 7 6 5

0, 3, 0, 2, 0,

0, 4, 0, 2, 3,

4 4, 3, 4, 5,

6, 6, 9 5, 5, 8, 9 5, 7 8

111

112

CH A P T E R 3: Using Statistics to Summarize Data Sets

FIGURE 3.6 Histogram of a bimodal data set.

By standing this ﬁgure on its side (or, equivalently, by turning the textbook), we can see that the corresponding histogram is approximately normal. Use it to assess the empirical rule. Solution A calculation yields that the sample mean and sample standard deviation of the data are x = 73.68 and

s = 12.80

The empirical rule states that approximately 68 percent of the data values are between x − s = 60.88 and x + s = 86.48. Since 17 of the observations actually fall within 60.88 and 86.48, the actual percentage is 100(17/25) = 68 percent. Similarly, the empirical rule states that approximately 95 percent of the data are between x − 2s = 48.08 and x + 2s = 96.28, whereas, in actuality, 100 percent of the data fall in this range. ■ A data set that is obtained by sampling from a population that is itself made up of subpopulations of different types is usually not normal. Rather, the histogram from such a data set often appears to resemble a combining, or superposition, of normal histograms and thus will often have more than one local peak or hump. Because the histogram will be higher at these local peaks than at their neighboring values, these peaks are similar to modes. A data set whose histogram has two local peaks is said to be bimodal. The data set represented in Fig. 3.6 is bimodal. Since a stem-and-leaf plot can be regarded as a histogram lying on its side, it is useful in showing us whether a data set is approximately normal. ■

Example 3.21 The following is the stem-and-leaf plot of the weights of 200 members of a health club.

3.6 Normal Data Sets and the Empirical Rule

24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8

9 1 7 2, 2, 5, 5, 6, 9, 9, 9 0, 0, 0, 0, 0, 1, 1, 2, 4, 4, 5, 8 0, 1, 1, 2, 2, 2, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6, 6, 7, 9, 9, 9 1, 1, 1, 2, 3, 3, 4, 4, 4, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, 9 0, 0, 1, 1, 1, 1, 2, 4, 5, 5, 6, 6, 8, 8, 8, 8 0, 1, 1, 1, 1, 1, 1, 5, 5, 5, 5, 6, 6, 6, 7, 7, 8, 9 0, 0, 0, 1, 2, 3, 4, 5, 6, 7, 7, 7, 8, 9, 9 0, 0, 0, 1, 1, 1, 2, 2, 2, 2, 2, 3, 3, 4, 5, 5, 6, 6, 6, 6, 7, 7, 8, 8, 8, 9, 9, 9 1, 1, 1, 2, 2, 2, 3, 4, 4, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 8, 8, 9, 9, 9 0, 1, 1, 2, 2, 2, 2, 3, 3, 4, 4, 5, 5, 6, 9, 9 0, 2, 3, 3, 3, 4, 4, 5, 7, 7, 8 0, 0, 9 6

By standing it on its side, we see that its histogram does not appear to be approximately normal. However, it is important to note that these data consist of the weights of all members of the health club, both female and male. Since these are clearly separate populations with regard to weight, it makes sense to consider the data for each gender separately. We will now do so. It turns out that these 200 data values are the weights of 97 women and 103 men. Separating the data for women and men results in the stem-and-leaf plots in Figs. 3.7 and 3.8. As we can see from the ﬁgures, the separated data for each sex appear to be approximately normal. Let us calculate x w , sw , x m , and sm , the sample mean and sample standard deviation for, respectively, the women and the men. This calculation yields x w = 125.70 sw = 15.58 16 15 14 13 12 11 10 9 8

x m = 174.69 sm = 21.23

0, 5 0, 1, 1, 1, 5 0, 0, 1, 2, 3, 4, 6, 7, 9 0, 0, 1, 1, 2, 2, 2, 2, 3, 4, 5, 5, 6, 6, 6, 6, 7, 8, 8, 8, 9, 9, 9 1, 1, 1, 2, 2, 2, 3, 4, 4, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 8, 8, 9, 9 0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 4, 4, 5, 5, 6, 9, 9 2, 3, 3, 3, 4, 4, 5, 7, 7, 8 0, 0, 9 6

FIGURE 3.7 Weights of 97 female health club members.

113

114

CH A P T E R 3: Using Statistics to Summarize Data Sets

24 23 22 21 20 19 18 17 16 15 14 13 12

9 1 7 2, 2, 5, 5, 6, 9, 9, 9 0, 0, 0, 0, 0, 1, 1, 2, 4, 4, 5, 8 0, 1, 1, 2, 2, 2, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6, 6, 7, 9, 9, 9 1, 1, 1, 2, 3, 3, 4, 4, 4, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, 9 0, 1, 1, 1, 1, 2, 4, 5, 6, 6, 8, 8, 8, 8 1, 1, 1, 5, 5, 5, 6, 6, 6, 7, 7, 8, 9 0, 5, 7, 7, 8, 9 0, 1, 2, 3, 7 9

FIGURE 3.8 Weights of 103 male health club members.

A further corroboration of the approximate normality of the two sets of separated data is provided by noting the similar values in each set of the sample mean and sample median. The sample median of the women’s weights is the 49th-smallest data value, which equals 126, whereas for the men’s data the sample median is the 52nd-smallest data value, which equals 174. These are quite close to the two sample means, whose values are 125.7 and 174.69. Given the values of the sample mean and sample standard deviation, it follows from the empirical rule that approximately 68 percent of the women will weigh between 110.1 and 141.3 and approximately 95 percent of the men will weigh between 132.2 and 217.2. The actual percentages from Figs. 3.7 and 3.8 are 100 ×

68 101 = 70.1 and 100 × = 98.1 97 103

■

PROBLEMS 1. The daily numbers of animals treated at a certain veterinarian clinic over a 24-day period are as follows: 22, 17, 19, 31, 28, 29, 21, 33, 36, 24, 15, 28, 25, 28, 22, 27, 33, 19, 25, 28, 26, 20, 30, 32 (a) Plot these data in a histogram. (b) Find the sample mean.

3.6 Normal Data Sets and the Empirical Rule

115

(c) Find the sample median. (d) Is this data set approximately normal?

Historical Perspective (North Wind Picture Archives)

Quetelet and How the Normal Curve Uncovered Fraud The Belgian social scientist and statistician Adolphe Quetelet was a great believer in the hypothesis that most data sets relating to human measurements are normal. In one study he measured the chests of 5738 Scottish soldiers, plotted the resulting data in a histogram, and concluded that it was normal. In a later study Quetelet used the shape of the normal histogram to uncover evidence of fraud in regard to draft conscripts to the French army. He studied data concerning the heights of a huge sample of 100,000 conscripts. Plotting the data in a histogram—with class intervals of 1 inch—he found that, with the exception of three class intervals around 62 inches, the data appeared to be normal. In particular, there were fewer values in the interval from 62 to 63 inches and slightly more in the intervals from 60 to 61 and from 61 to 62 inches than would have occurred with a perfect normal ﬁt of the data. Trying to ﬁgure out why the normal curve did not ﬁt as well as he had supposed it would, Quetelet discovered that 62 inches was the minimum height required for soldiers in the French army. Based on this and his conﬁdence in the widespread applicability of normal data, Quetelet concluded that some conscripts whose heights were slightly above 62 inches were “bending their knees” to appear shorter so as to avoid the draft. For 50 years following Quetelet, that is, roughly from 1840 to 1890, it was widely believed that most data sets from homogeneous populations (that is, data that were not obviously a mixture of different populations) would appear to be normal if the sample size were sufﬁciently large. Whereas present-day statisticians have become somewhat skeptical about this claim, it is quite common for a data set to appear to come from a normal population. This phenomenon, which often appears in data sets originating in either the biological or the physical sciences, is partially explained by a mathematical result known as the central limit theorem. Indeed, the central limit theorem (studied in Chap. 7) will in itself explain why many data sets originating in the physical sciences are approximately normal. To explain why biometric data (that is, data generated by studies in biology) often appear to be normal, we will use what was originally an empirical observation noted by Francis Galton but that nowadays has a sound scientiﬁc explanation, called regression to the mean. Regression to the mean, in conjunction with the central limit theorem and the passing of many generations, will yield our explanation as to why a biometric data set is often normal. The explanation will be presented in Chap. 12.

Adolphe Quetelet

116

CH A P T E R 3: Using Statistics to Summarize Data Sets

2. The following data give the injury rates per 100,000 worker-hours for a sample of 20 semiconductor ﬁrms: 1.4, 2.4, 3.7, 3.1, 2.0, 1.9, 2.5, 2.8, 2.2, 1.7, 3.1, 4.0, 2.2, 1.8, 2.6, 3.6, 2.9, 3.3, 2.0, 2.4 (a) (b) (c) (d)

Plot the data in a histogram. Is the data set roughly symmetric? If the answer to (b) is no, is it skewed to the left or to the right? If the answer to (b) is yes, is it approximately normal?

The following table gives the 2006 per capita consumption of milk in various countries. Problems 3 and 4 refer to this table. Per Capita Consumption of Milk and Milk Products in Various Countries, 2006 data Country Finland Sweden Ireland Netherlands Norway Spain (2005) Switzerland United Kingdom (2005) Australia (2005) Canada (2005) European Union (25 countries) Germany France New Zealand (2005) United States Austria Greece Argentina (2005) Italy Mexico China (2005)

Liquid milk drinks (litres) 183.9 145.5 129.8 122.9 116.7 119.1 112.5 111.2 106.3 94.7 92.6 92.3 92.2 90.0 83.9 80.2 69.0 65.8 57.3 40.7 8.8

Source: International Dairy Federation, Bulletin 423/2007.

3. Find the sample mean and sample median of the milk consumption data set. 4. Plot the milk consumption data in a stem and leaf plot. Is the data set approximately normal?

3.6 Normal Data Sets and the Empirical Rule

5. The following represent the times (in minutes) it took 22 newly hired workers to complete a standardized task: 166, 82, 175, 181, 169, 177, 180, 185, 159, 164, 170, 149, 188, 173, 170, 164, 158, 177, 173, 175, 190, 172 (a) Find the sample mean. (b) Find the sample median. (c) Plot the data in a histogram. (d) Is this data set approximately normal? 6. The following data give the age at inauguration of all 43 presidents of the United States.

President 1. Washington 2. J. Adams 3. Jefferson 4. Madison 5. Monroe 6. J.Q. Adams 7. Jackson 8. Van Buren 9. W. Harrison 10. Tyler 11. Polk 12. Taylor 13. Fillmore 14. Pierce 15. Buchanan 16. Lincoln 17. A. Johnson 18. Grant 19. Hayes 20. Garfield 21. Arthur 22. Cleveland

Age at inauguration 57 61 57 57 58 57 61 54 68 51 49 64 50 48 65 52 56 46 54 49 50 47

President 23. B. Harrison 24. Cleveland 25. McKinley 26. T. Roosevelt 27. Taft 28. Wilson 29. Harding 30. Coolidge 31. Hoover 32. F. Roosevelt 33. Truman 34. Eisenhower 35. Kennedy 36. L. Johnson 37. Nixon 38. Ford 39. Carter 40. Reagan 41. G. H. W. Bush 42. Clinton 43. G. W. Bush 44. Obama

Age at inauguration 55 55 54 42 51 56 55 51 54 51 60 62 43 55 56 61 52 69 64 46 54 47

(a) Find the sample mean and sample standard deviation of this data set. (b) Draw a histogram for the given data. (c) Do the data appear to be approximately normal?

117

118

CH A P T E R 3: Using Statistics to Summarize Data Sets

(d) If the answer to (c) is yes, give an interval that you would expect to contain approximately 95 percent of the data observations. (e) What percentage of the data lies in the interval given in part (d)? 7. For the data on the weights of female health club members presented in Fig. 3.7, the sample mean and sample standard deviation were computed to be 125.70 and 15.58, respectively. Based on the shape of Fig. 3.7 and these values, approximate the proportion of the women whose weight is between 94.54 and 156.86 pounds. What is the actual proportion? 8. A sample of 36 male coronary patients yielded the following data concerning the ages at which they suffered their ﬁrst heart attacks. 7 6 5 4 3

1, 2, 4, 5 0, 1, 2, 2, 3, 4, 5, 7 0, 1, 2, 3, 3, 4, 4, 4, 5, 6, 7, 8, 9 1, 2, 2, 3, 4, 5, 7, 8, 9 7, 9

(a) Determine x and s. (b) From the shape of the stem-and-leaf plot, what percentage of data values would you expect to be between x − s and x + s? Between x − 2s and x + 2s? (c) Find the actual percentages for the intervals given in (b). 9. If the histogram is skewed to the right, which statistic will be larger— the sample mean or the sample median? (Hint : If you are not certain, construct a data set that is skewed to the right and then calculate the sample mean and sample median.) 10. The following data are the ages of a sample of 36 victims of violent crime in a large eastern city: 25, 16, 14, 22, 17, 20, 15, 18, 33, 52, 70, 38, 18, 13, 22, 27, 19, 23, 33, 15, 13, 62, 21, 57, 66, 16, 24, 22, 31, 17, 20, 14, 26, 30, 18, 25 (a) (b) (c) (d) (e)

Determine the sample mean. Find the sample median. Determine the sample standard deviation. Does this data set appear to be approximately normal? What proportion of the data lies within 1 sample standard deviation of the sample mean? (f) Compare your answer in (e) to the approximation provided by the empirical rule.

The following table lists the 2002 per capita income for the 50 states. Problems 11 to 13 refer to it.

3.6 Normal Data Sets and the Empirical Rule

11. Using the data on the ﬁrst 25 states, (a) Plot the data in a histogram. (b) Compute the sample mean. (c) Compute the sample median. (d) Compute the sample variance. (e) Are the data approximately normal? (f) Use the empirical rule to give an interval which should contain approximately 68 percent of the observations. (g) Use the empirical rule to give an interval which should contain approximately 95 percent of the observations. (h) Determine the actual proportion of observations in the interval speciﬁed in (f). (i) Determine the actual proportion of observations in the interval speciﬁed in (g). 12. Repeat Prob. 11, this time using the data on the ﬁnal 25 states. 13. Repeat Prob. 11, this time using all the data in the table. Personal Income per Capita in Constant (1996) Dollars, 2002 State

Income

Rank

United States Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Hawaii Idaho Illinois Indiana Iowa New York North Carolina North Dakota Ohio Oklahoma Oregon

27,857 22,624 28,947 23,573 21,169 29,707 29,959 38,450 29,512 26,646 25,949 27,011 22,560 30,075 25,425 25,461 32,451 24,949 24,293 26,474 23,026 25,867

(X) 43 14 38 49 10 9 1 12 23 28 20 44 8 32 31 5 34 36 25 40 29

State

Income

Rank

Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico Tennessee Texas Utah Vermont Virginia Washington

26,237 23,030 22,910 24,979 32,680 35,333 27,276 30,675 20,142 26,052 22,526 26,804 27,172 30,912 35,521 21,555 24,913 25,705 21,883 26,620 29,641 29,420

26 39 41 33 4 3 18 7 50 27 45 22 19 6 2 47 35 30 46 24 11 13 (Continued)

119

120

CH A P T E R 3: Using Statistics to Summarize Data Sets

(Continued) State

Income

Rank

Pennsylvania Rhode Island South Carolina South Dakota

28,565 28,198 22,868 24,214

15 16 42 37

State

Income

Rank

West Virginia Wisconsin Wyoming

21,327 26,941 27,530

48 21 17

Note: When states share the same rank, the next lower rank is omitted. Because of rounded data, states may have identical values shown, but different ranks.

3.7 SAMPLE CORRELATION COEFFICIENT Consider the data set of paired values (x1 , y1 ), (x2 , y2 ), . . . , (xn , yn ). In this section we will present a statistic, called the sample correlation coefﬁcient, that measures the degree to which larger x values go with larger y values and smaller x values go with smaller y values. The data in Table 3.3 represent the average daily number of cigarettes smoked (the x variable) and the number of free radicals (the y variable), in a suitable unit, found in the lungs of 10 smokers. (A free radical is a single atom of oxygen. It is believed to be potentially harmful because it is highly reactive and has a strong tendency to combine with other atoms within the body.) Figure 3.9 shows the scatter diagram for these data. From an examination of Fig. 3.9 we see that when the number of cigarettes is high, there tends to be a large number of free radicals, and when the number of

Table 3.3 Cigarette Smoking and Free Radicals Person 1 2 3 4 5 6 7 8 9 10

Number of cigarettes smoked

Free radicals

18 32 25 60 12 25 50 15 22 30

202 644 411 755 144 302 512 223 183 375

3.7 Sample Correlation Coefﬁcient

FIGURE 3.9 Cigarettes smoked versus number of free radicals.

Table 3.4 Pulse Rate and Years of School Completed Person 1

2

3

4

5

6

7

8

9

10

Years of school

12

16

13

18

19

12

18

19

12

14

Pulse rate

73

67

74

63

73

84

60

62

76

71

cigarettes smoked is low, there tends to be a small number of free radicals. In this case, we say that there is a positive correlation between these two variables. We are also interested in determining the strength of the relationship between a pair of variables in which large values of one variable tend to be associated with small values of the other. For instance, the data of Table 3.4 represent the years of schooling (variable x) and the resting pulse rate in beats per minute (variable y) of 10 individuals. A scatter diagram of this data is presented in Fig. 3.10. From Fig. 3.10 we see that higher numbers of years of schooling tend to be associated with lower resting pulse rates and that lower numbers of years of schooling tend to be associated with the higher resting pulse rates. This is an example of a negative correlation. To obtain a statistic that can be used to measure the association between the individual values of a paired set, suppose the data set consists of the paired values (xi , yi ), i = 1, . . . , n. Let x and y denote the sample mean of the x values and the sample mean of the y values, respectively. For data pair i, consider xi − x the

121

122

CH A P T E R 3: Using Statistics to Summarize Data Sets

FIGURE 3.10 Scatter diagram of years in school and pulse rate.

deviation of its x value from the sample mean and yi − y the deviation of its y value from the sample mean. Now if xi is a large x value, then it will be larger than the average value of all the x’s and so the deviation xi − x will be a positive value. Similarly, when xi is a small x value, then the deviation xi − x will be a negative value. Since the same statements are true about the y deviations, we can conclude the following. When large values of the x variable tend to be associated with large values of the y variable and small values of the x variable tend to be associated with small values of the y variable, then the signs, either positive or negative, of xi − x and yi − y will tend to be the same. Now, if xi − x and yi − y both have the same sign (either positive or negative), then their product (xi − x)(yi − y) will be positive. Thus, it follows that when large x values tend to be associated with large y values and small x values are associated with small y values, then ni=1 (xi − x)(yi − y) will tend to be a large positive number. The same logic also implies that when large values of one of the variables tend to go along with small values of the other, then the signs of xi − x and yi − y will be opposite, and so ni=1 (xi − x)(yi − y) will be a large negative number. To determine what it means for ni=1 (xi − x)(yi − y) to be “large,” we standardize this sum ﬁrst by dividing by n − 1 and then by dividing by the product of the two sample standard deviations. The resulting statistic is called the sample correlation coefﬁcient.

3.7 Sample Correlation Coefﬁcient

Deﬁnition Let sx and sy denote, respectively, the sample standard deviations of the x values and the y values. The sample correlation coefﬁcient, call it r, of the data pairs (xi , yi ), i = 1, . . . , n, is deﬁned by n (xi − x)(yi − y) r = i=1 (n − 1)sx sy n

=

(xi − x)(yi − y)

i=1 n

(xi − x)2

i=1

n

(yi − y)2

i=1

When r > 0, we say that the sample data pairs are positively correlated; and when r < 0, we say that they are negatively correlated. We now list some of the properties of the sample correlation coefﬁcient. 1. The sample correlation coefﬁcient r is always between −1 and +1. 2. The sample correlation coefﬁcient r will equal +1 if, for some constant a, yi = a + bxi

i = 1, . . . , n

where b is a positive constant. 3. The sample correlation coefﬁcient r will equal −1 if, for some constant a, yi = a + bxi

i = 1, . . . , n

where b is a negative constant. 4. If r is the sample correlation coefﬁcient for the data xi , yi , i = 1, . . . , n, then for any constants a, b, c, d, it is also the sample correlation coefﬁcient for the data a + bxi , c + dyi

i = 1, . . . , n

provided that b and d have the same sign (that is, provided that bd ≥ 0). Property 1 says that the sample correlation coefﬁcient r is always between −1 and +1. Property 2 says that r will equal +1 when there is a straight-line (also called a linear) relation between the paired data such that large y values are attached to large x values. Property 3 says that r will equal −1 when the relation is linear and large y values are attached to small x values. Property 4 states that the value of r is unchanged when a constant is added to each of the x variables (or to each of the y variables) or when each x variable (or each y variable) is multiplied by a positive constant. This property implies that r does not depend on the dimensions chosen to measure the data. For instance, the sample correlation coefﬁcient between a person’s height and weight does not depend on whether the height is measured in feet or in inches or whether the weight is measured in pounds or kilograms. Also if one of the values in the pair is temperature, then the sample correlation coefﬁcient is the same whether it is measured in degrees Fahrenheit or Celsius.

123

124

CH A P T E R 3: Using Statistics to Summarize Data Sets

For computational purposes, the following is a convenient formula for the sample correlation coefﬁcient. Computational Formula for r n

r =

n i=1

■

xi yi − nxy

i=1

xi2

− nx

2

n i=1

yi2

− ny

2

Example 3.22 The following table gives the U.S. per capita consumption of whole milk (x) and of low-fat milk (y) in three different years. Per capita consumption (gallons) 1980

1984

1984

Whole milk (x )

17.1

14.7

12.8

Low-fat milk (y )

10.6

11.5

13.2

Source: U.S. Department of Agriculture, Food Consumption, Prices, and Expenditures.

Find the sample correlation coefﬁcient r for the given data. Solution To make the computation easier, let us ﬁrst subtract 12.8 from each of the x values and 10.6 from each of the y values. This gives the new set of data pairs: i 1

2

3

xi

4.3

1.9

0

yi

0

0.9

2.6

Now, x=

4.3 + 1.9 + 0 = 2.0667 3

y=

0 + 0.9 + 2.6 = 1.1667 3

3.7 Sample Correlation Coefﬁcient

3

xi yi = (1.9)(0.9) = 1.71

i=1 3

xi2 = (4.3)2 + (1.9)2 = 22.10

i=1 3

yi2 = (0.9)2 + (2.6)2 = 7.57

i=1

Thus, 1.71 − 3(2.0667)(1.1667) = −0.97 [22.10 − 3(2.0667)2 ][7.57 − 3(1.1667)2 ]

r=√

Therefore, our three data pairs exhibit a very strong negative correlation between consumption of whole and of low-fat milk. For small data sets such as in Example 3.22, the sample correlation coefﬁcient can be easily obtained by hand. However, for large data sets this computation can become tedious, and a calculator or statistical software is useful. ■ ■

Example 3.23 Compute the sample correlation coefﬁcient of the data of Table 3.3, which relates the number of cigarettes smoked to the number of free radicals found in a person’s lungs. Solution The number of pairs is 10. The pairs are as follows: 18, 202 32, 644 25, 411 60, 755 12, 144 25, 302 50, 512 15, 223 22, 183 30, 375 A calculation shows that the sample correlation coefﬁcient is 0.8759639.

■

125

126

CH A P T E R 3: Using Statistics to Summarize Data Sets

The large value of the sample correlation coefﬁcient indicates a strong positive correlation between the number of cigarettes a person smokes and the number of free radicals in that person’s lungs.

■

Example 3.24 Compute the sample correlation coefﬁcient of the data of Table 3.4, which relates a person’s resting pulse rate to the number of years of school completed. Solution The pairs are as follows: 12, 73 16, 67 13, 74 18, 63 19, 73 12, 84 18, 60 19, 62 12, 76 14, 71 The sample correlation coefﬁcient is −0.763803. The large negative value of the sample correlation coefﬁcient indicates that, for the data set considered, a high pulse rate tends to be associated with a small number of years spent in school and a low pulse rate tends to be associated with a large number of years spent in school. ■

The absolute value of the sample correlation coefﬁcient r (that is, |r|—its value without regard to its sign) is a measure of the strength of the linear relationship between the x and the y values of a data pair. A value of |r| equal to 1 means that there is a perfect linear relation; that is, a straight line can pass through all the data points (xi , yi ), i = 1, . . . , n. A value of |r| of about 0.8 means that the linear relation is relatively strong; although there is no straight line that passes through all the data points, there is one that is “close” to them all. A value of |r| around 0.3 means that the linear relation is relatively weak. The sign of r gives the direction of the relation. It is positive when the linear relation is such that smaller y values tend to

3.7 Sample Correlation Coefﬁcient

FIGURE 3.11 Sample correlation coefﬁcients.

go with smaller x values and larger y values with larger x values (and so a straightline approximation points upward); and it is negative when larger y values tend to go with smaller x values and smaller y values with larger x values (and so a straightline approximation points downward). Figure 3.11 displays scatter diagrams for data sets with various values of r.

The development of the concept and utility of the sample correlation coefﬁcient involved the efforts of four of the great men of statistics. The original concept was due to Francis Galton, who was trying to study the laws of inheritance from a quantitative point of view. As such, he wanted to be able to quantify the degree to which characteristics of an offspring relate to those of its parents. This led him to name and deﬁne a form of the sample correlation coefﬁcient that differs somewhat from the one presently in use. Although originally it was meant to be used to assess the hereditary inﬂuence of a parent on an offspring, Galton later realized that the sample correlation coefﬁcient presented a method of assessing the interrelation between any two variables. Although Francis Galton was the founder of the ﬁeld of biometrics—the quantitative study of biology—its acknowledged leader, at least after 1900, was Karl Pearson. After the Royal Society of London passed a resolution in 1900 stating that it would no longer accept papers that applied mathematics to the study

(Bettmann)

Historical Perspective

Francis Galton

127

128

CH A P T E R 3: Using Statistics to Summarize Data Sets

of biology, Pearson, with ﬁnancial assistance from Galton, founded the statistical journal Biometrika, which still ﬂourishes today. The form of the sample correlation coefﬁcient that is presently in use (and that we have presented) is due to Karl Pearson and was originally called Pearson’s product-moment correlation coefﬁcient. The probabilities associated with the possible values of the sample correlation coefﬁcient r were discovered, in the case where the data pairs come from a normal population, by William Gosset. There were, however, some technical errors in his derivations, and these were subsequently corrected in a paper by Ronald Fisher.

PROBLEMS 1. Explain why the sample correlation coefﬁcient for the data pairs (121, 360), (242, 362), (363, 364) is the same as that for the pairs (1, 0), (2, 2), (3, 4) which is the same as that for the pairs (1, 0), (2, 1), (3, 2) 2. Compute the sample correlation coefﬁcient for the data pairs in Prob. 1.

Statistics In Perspective Correlation Measures Association, Not Causation The results of Example 3.24 indicated a strong negative correlation between an individual’s years of education and that individual’s resting pulse rate. However, this does not imply that additional years of school will directly reduce one’s pulse rate. That is, whereas additional years of school tend to be associated with a lower resting pulse rate, this does not mean that it is a direct cause of it. Often the explanation for such an association lies with an unexpressed factor that is related to both variables under consideration. In this instance, it may be that a person who has spent additional time in school is more aware of the latest ﬁndings in the area of health and thus may be more aware of the importance of exercise and good nutrition; or perhaps it is not knowledge that is making the difference but rather that people who have had more education tend to end up in jobs that allow them more time for exercise and good nutrition. Probably the strong negative correlation between years in school and resting pulse rate results from a combination of these as well as other underlying factors.

3.7 Sample Correlation Coefﬁcient

3. The following data represent the IQ scores of 10 mothers and their eldest daughters. Mother’s IQ

Daughter’s IQ

135 127 124 120 115 112 104 96 94 85

121 131 112 115 99 118 106 89 92 90

(a) (b) (c) (d)

Draw a scatter diagram. Guess at the value of the sample correlation coefﬁcient r. Compute r. What conclusions can you draw about the relationship between the mother’s and daughter’s IQs? 4. The following is a sampling of 10 recently released ﬁrst-time federal prisoners. The data give their crime, their sentence, and the actual time that they served. Number

Crime

1 2 3 4 5 6 7 8 9 10

Drug abuse Forgery Drug abuse Kidnapping Income tax fraud Drug abuse Robbery Embezzlement Robbery Robbery

Sentence (months)

Time served (months)

44 30 52 240 18 60 120 24 60 96

24 12 26 96 12 28 52 14 35 49

Draw a scatter diagram of the sentence time versus time actually served. Compute the sample correlation coefﬁcient. What does this say about the relationship between the length of a sentence and the time actually served? 5. Using the data of Prob. 4, determine the sample correlation coefﬁcient of the sentence time and the proportion of that time actually served.

129

130

CH A P T E R 3: Using Statistics to Summarize Data Sets

What does this say about the relationship between the length of a sentence and the proportion of this time that is actually served? 6. The following data refer to the number of adults in prison and on parole in 12 midwestern states. The data are in thousands of adults. State Illinois Indiana Iowa Kansas Michigan Minnesota Missouri Nebraska North Dakota Ohio South Dakota Wisconsin

In prison

On parole

18.63 9.90 2.83 4.73 17.80 2.34 9.92 1.81 0.42 20.86 1.05 5.44

11.42 2.80 1.97 2.28 6.64 1.36 4.53 0.36 0.17 6.51 0.42 3.85

(a) Draw a scatter diagram. (b) Determine the sample correlation coefﬁcient between the number of adults in state prison and on parole in that state. (c) Fill in the missing word. States having a large prison population tend to have a(n)_____number of individuals on parole. 7. The following data relate the number of criminal cases ﬁled in various U.S. cities to the percentage of those cases that result in a plea of guilty.

City San Diego, CA Dallas, TX Portland, OR Chicago, IL Denver, CO Philadelphia, PA Lansing, MI St. Louis, MO Davenport, IA Tallahassee, FL Salt Lake City, UT

Percentage of cases resulting in a guilty plea

Number of cases ﬁled

73 72 62 41 68 26 68 63 60 50 61

11,534 14,784 3,892 35,528 3,772 13,796 1,358 3,649 1,312 2,879 2,745

Determine the sample correlation coefﬁcient between the number of cases ﬁled and the percentage of guilty pleas. What can you say about the degree of association between these two variables for these data?

3.7 Sample Correlation Coefﬁcient

8. The following table gives yearly per capita soft drink consumption (in litres) and the yearly per capita milk consumption (in kg) for a variety of countries. Use it to ﬁnd the sample correlation coefﬁcient between soft drink and milk consumption. Per capita soft drink and milk consumption Country United States Australia Switzerland France United Kingdom The Netherlands New Zealand Germany Italy Japan

soft drink

milk

216 100 81 37 97 96 84 72 50 22

254 233 308 256 230 329 210 314 239 68

9. The following table lists per capita income data both for the U.S. and for residents of the state of Colorado for each of the years from 1992 to 2007. Use it to compute the sample correlation coefﬁcient between U.S. and Colorado per capita income. Annual Per Capita Personal Income

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

United States

Colorado

$20,854 $21,346 $22,172 $23,076 $24,175 $25,334 $26,883 $27,939 $29,845 $30,574 $30,821 $31,504 $33,123 $34,757 $36,714 $38,611

$21,109 $22,054 $23,004 $24,226 $25,570 $26,846 $28,784 $30,492 $33,361 $34,438 $33,956 $33,989 $35,523 $37,600 $39,491 $41,042

131

132

CH A P T E R 3: Using Statistics to Summarize Data Sets

10. The following data give the numbers of physicians and dentists, per 100,000 population, in the United States for six different years.

Physicians Dentists

1980

1981

1982

1983

1985

1986

2001

211 54

217 54

222 55

228 56

237 57

246 57

253 59

Source: Health Resources Statistics, annual.

(a) Show that the number of physicians and the number of dentists are positively correlated for these years. (b) Do you think that a large value of one of these variables by itself causes a large value of the other? If not, how would you explain the reason for the positive correlation? The following table gives the death rates by selected causes in different countries. It will be used in Probs. 11 to 13.

Death Rates per 100,000 Population by Selected Causes and Countries Malignant neoplasm of—

Country

Year

Ischemic heart disease

United States Australia Austria Belgium Bulgaria Canada Czechoslovakia Denmark Finland France Hungary Italy Japan Netherlands New Zealand Norway Poland

1984 1985 1986 1984 1985 1985 1985 1985 1986 1985 1986 1983 1986 1985 1985 1985 1986

218.1 230.9 155.1 120.6 245.9 200.6 289.4 243.8 259.8 76.0 240.1 128.9 41.9 164.6 250.5 208.5 109.4

Cerebrovascular disease

Lung, trachea, bronchus

Stomach

60.1 95.6 133.2 95.0 254.5 57.5 194.3 73.4 105.0 79.7 186.5 121.9 112.8 71.1 98.4 88.6 75.3

52.7 41.0 34.3 55.9 30.6 50.6 51.3 52.2 36.4 32.2 55.0 42.1 24.9 56.3 42.0 26.3 47.2

6.0 10.1 20.7 14.7 24.2 9.0 22.4 10.9 17.3 10.8 25.9 23.9 40.7 15.6 11.2 14.4 24.2

Female breast

Bronchitis, emphysema, asthma

Chronic lever disease and cirrhosis

31.9 30.0 31.6 36.8 21.5 34.5 27.3 39.7 23.9 27.1 31.2 28.9 8.1 38.2 37.7 25.9 21.1

8.3 16.9 22.3 22.6 28.6 9.7 33.8 37.1 19.8 11.7 43.8 30.9 12.2 17.8 25.8 18.2 33.4

12.9 8.7 26.6 12.4 16.2 10.1 19.6 12.2 8.8 22.9 42.1 31.5 14.4 5.5 4.8 6.9 12.0

3.7 Sample Correlation Coefﬁcient

(Continued ) Malignant neoplasm of— Cerebrovascular disease

Lung, trachea, bronchus

Stomach

Female breast

Bronchitis, emphysema, asthma

Chronic lever disease and cirrhosis

Country

Year

Ischemic heart disease

Portugal Spain Sweden Switzerland United Kingdom: England and Wales Scotland West Germany

1986 1981 1985 1986

76.6 79.0 244.7 112.0

216.4 133.9 73.0 65.6

18.7 26.0 23.2 36.6

26.5 19.7 12.5 12.0

22.6 19.0 26.0 36.6

17.8 19.1 14.3 17.5

30.0 23.3 6.4 10.4

1985 1986 1986

247.6 288.0 159.5

104.5 128.4 100.4

57.2 68.7 34.6

15.2 14.9 18.3

41.9 41.2 32.6

24.2 14.8 26.1

4.8 7.3 19.3

Source: World Health Organization, World Health Statistics.

In doing Probs. 11 to 13, use all the data if you are running either Program 3-2 or a statistical package. If you are working with a hand calculator, use only the data relating to the ﬁrst seven countries. 11. Find the sample correlation coefﬁcient between the death rates of ischemic heart disease and of chronic liver disease. 12. Find the sample correlation coefﬁcient between the death rates of stomach cancer and of female breast cancer. 13. Find the sample correlation coefﬁcient between the death rates of lung cancer and of bronchitis, emphysema, and asthma. 14. In a well-publicized experiment, a University of Pittsburgh researcher enlisted the cooperation of public school teachers in Boston in obtaining a baby tooth from each of their pupils. These teeth were then sawed open and analyzed for lead content. The lead content of each tooth was plotted against the pupil’s IQ test score. A strong negative correlation resulted between the amount of lead in the teeth and the IQ scores. Newspapers headlined this result as “proof” that lead ingestion results in decreased scholastic aptitude. (a) Does this conclusion necessarily follow? (b) Offer some other possible explanations. 15. A recent study has found a strong positive correlation between the cholesterol levels of young adults and the amounts of time they spend watching television. (a) Would you have expected such a result? Why? (b) Do you think that watching television causes higher cholesterol levels?

133

134

CH A P T E R 3: Using Statistics to Summarize Data Sets

16.

17.

18.

19.

20.

(c) Do you think that having a high cholesterol level makes a young adult more likely to watch television? (d) How would you explain the results of the study? An analysis relating the number of points scored and fouls committed by basketball players in the Paciﬁc Ten conference has established a strong positive correlation between these two variables. The analyst has gone on record as claiming that this veriﬁes the hypothesis that offensive-minded basketball players tend to be very aggressive and so tend to commit a large number of fouls. Can you think of a simpler explanation for the positive correlation? (Hint: Think in terms of the average number of minutes per game that a player is on the court.) A New England Journal of Medicine study published in October 1993 found that people who have guns in their homes for protection are 3 times more likely to be murdered than those with no guns in the home. Does this prove that an individual’s chance of being murdered is increased when he or she purchases a gun to keep at home? Explain your answer. If for each of the ﬁfty states we plot the paired data consisting of the average income of residents of the state and the number of foreignborn immigrants who reside in the state, then the data pairs will have a positive correlation. Can we conclude that immigrants tend to have higher incomes than native-born Americans? If not, how else could this phenomenon be explained? A recent study (reported in the May 5, 2008 LA Times) yielded a positive correlation between breast-fed babies and scores on a vocabulary test taken at age 6. Discuss the potential difﬁculties in interpreting the results of this study. A recent study (reported in the March 10, 2009 NY Times) yielded a negative correlation between the age of the father and the results of cognitive tests given to the infant at ages 8 months, 4 years old, and 7 years old. Although the differences in scores between those infants having younger and older fathers was slight the authors of the study called the ﬁndings “unexpectedly startling.” (On the other hand there was a positive correlation between a mother’s age and the cognitive test scores.) Discuss the potential difﬁculties in interpreting the results of this study.

KEY TERMS Statistic: A numerical quantity whose value is determined by the data. Sample mean: The arithmetic average of the values in a data set.

Key Terms

Deviation: The difference between the individual data values and the sample mean. If xi is the ith data value and x is the sample mean, then xi − x is called the ith deviation. Sample median: The middle value of an ordered set of data. For a data set of n values, the sample median is the (n + 1)/2-smallest value when n is odd and the average of the n/2- and n/2 + 1-smallest values when n is even. Sample 100p percentile: That data value such that at least 100p percent of the data are less than or equal to it and at least 100(1 − p) percent of the data are greater than or equal to it. If two data values satisfy this criterion, then it is the average of them. First quartile: The sample 25th percentile. Second quartile: The sample 50th percentile, which is also the sample median. Third quartile: The sample 75th percentile. Sample mode: The data value that occurs most frequently in a data set. Sample variance: The statistic s2 , deﬁned by n s2 =

i=1 (xi

− x)2 n−1

It measures the average of the squared deviations. Sample standard deviation: The positive square root of the sample variance. Range: The largest minus the smallest data value. Interquartile range: The third quartile minus the ﬁrst quartile. Normal data set: One whose histogram is symmetric about its middle interval and decreases on both sides of the middle in a bell-shaped manner. Skewed data set: One whose histogram is not symmetric about its middle interval. It is said to be skewed to the right if it has a long tail to the right and skewed to the left if it has a long tail to the left. Bimodal data set: One whose histogram has two local peaks or humps. Sample correlation coefﬁcient: For the set of paired values xi , yi , i = 1, . . . , n, it is deﬁned by n r=

i=1 (xi

− x)(yi − y) (n − 1)sx sy

where x and sx are, respectively, the sample mean and the sample standard deviation of the x values, and similarly for y and sy . A value of r near +1 indicates that larger x values tend to be paired with larger y values and smaller x values tend to be paired with smaller y values. A value near −1 indicates that larger x values

135

136

CH A P T E R 3: Using Statistics to Summarize Data Sets

tend to be paired with smaller y values and smaller x values tend to be paired with larger y values.

SUMMARY We have seen three different statistics which describe the center of a data set: the sample mean, sample median, and sample mode. The sample mean of the data x1 . . . , xn is deﬁned by n xi x = i=1 n and is a measure of the center of the data. If the data are speciﬁed by the frequency table Value

Frequency

x1 x2 .. .

f1 f2 .. .

xk

fk

then the sample mean of the n =

k

i=1 fi

data values can be expressed as

n x=

i=1 fi xi

n

A useful identity is n

(xi − x) = 0

i=1

The sample median is the middle value when the data are arranged from smallest to largest. If there are an even number of data points, then it is the average of the two middle values. It is also a measure of the center of the data set. The sample mode is that value in the data set that occurs most frequently. Suppose a data set of size n is arranged from smallest to largest. If np is not an integer, then the sample 100p percentile is the value whose position is the smallest integer larger than np. If np is an integer, then the sample 100p percentile is the average of the values in positions np and np + 1.

Summary

The sample 25th percentile is the ﬁrst quartile. The sample 50th percentile (which is equal to the sample median) is called the second quartile, and the sample 75th percentile is called the third quartile. The sample variance s2 is a measure of the spread in the data and is deﬁned by n (xi − x)2 2 s = i=1 n−1 where n is the size of the set. Its square root s is called the sample standard deviation, and it is measured in the same units as the data. The following identity is useful for computing the sample variance by using pencil and paper or a hand calculator. n i=1

(xi − x)2 =

n

xi2 − nx 2

i=1

Program 3-1 will compute the sample mean, sample variance, and sample standard deviation of any set of data. Another statistic that describes the spread of the data is the range, the difference between the largest and smallest data values.

137

138

CH A P T E R 3: Using Statistics to Summarize Data Sets

Normal data sets will have their sample mean and sample median approximately equal. Their histograms are symmetric about the middle interval and exhibit a bell shape. The sample correlation coefﬁcient r measures the degree of association between two variables. Its value is between −1 and +1. A value of r near +1 indicates that when one of the variables is large, the other one also tends to be large and that when one of them is small, the other also tends to be small. A value of r near −1 indicates that when one of the variables is large, the other one tends to be small. A large value of |r| indicates a strong association between the two variables. Association does not imply causation.

REVIEW PROBLEMS 1. Construct a data set that is symmetric about 0 and contains (a) Four distinct values (b) Five distinct values (c) In both cases, compute the sample mean and sample median. 2. The following stem-and-leaf plot records the diastolic blood pressure of a sample of 30 men. 9 8 7 6 5

3, 5, 8 6, 7, 8, 9, 9, 9 0, 1, 2, 2, 4, 5, 5, 6, 7, 8 0, 1, 2, 2, 3, 4, 5, 5 4, 6, 8

(a) (b) (c) (d) (e) (f) (g)

Compute the sample mean x. Compute the sample median. Compute the sample mode. Compute the sample standard deviation s. Do the data appear to be approximately normal? What proportion of the data values lies between x + 2s and x − 2s? Compare the answer in part (f) to the one prescribed by the empirical rule. 3. The following data are the median ages of residents in each of the 50 states of the United States: 29.3 27.7 30.4 31.1 28.5 32.1 28.0 31.3 26.6 25.8 25.9 33.0 31.5 30.0 28.4 24.9 31.6 26.6 25.4 29.2 29.3 27.9 31.8 31.5 30.3

Review Problems

28.5 29.3 26.6 31.2 32.1 31.4 30.1 27.0 28.5 27.6 28.9 29.4 30.5 31.2 29.4 29.3 30.1 28.8 27.9 30.4 32.3 30.4 25.8 27.1 26.9 (a) Find the median of these ages. (b) Is this necessarily the median age of all people in the United States? Explain. (c) Find the quartiles. (d) Find the sample 90th percentile. 4. Use Table 3.2 in Example 3.19 to ﬁll in the answers. (a) To have one’s score be among the top 10 percent of all physical science students, it must be at least___. (b) To have one’s score be among the top 25 percent of all social science students, it must be at least___. (c) To have one’s score be among the bottom 50 percent of all medical students, it must be less than or equal to___. (d) To have one’s score be among the middle 50 percent of all law school students, it must be between___ and___. 5. The number of violent offenses per 100,000 population is given here for each of the 50 states. Is this data set approximately normal?

Violent Crime per 100,000 Population, 2002 State

Rate

Rank

United States Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Hawaii Indiana Idaho

495 444 563 553 424 593 352 311 599 770 459 262 357 255

(X) 21 12 13 22 10 27 33 9 2 20 41 26 42

State

Rate

Rank

Illinois Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana Nebraska

621 286 377 279 662 108 770 484 540 268 343 539 352 314

8 36 24 38 6 48 2 18 14 40 31 15 27 32 (Continued )

139

140

CH A P T E R 3: Using Statistics to Summarize Data Sets

(Continued) State

Rate

Rank

Nevada New Hampshire New Jersey New Mexico New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island

638 161 375 740 496 470 78 351 503 292 402 285

7 47 25 4 17 19 50 29 16 34 23 37

State

Rate

Rank

South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming

822 177 717 579 237 107 291 345 234 225 274

1 46 5 11 43 49 35 30 44 45 39

Note: Violent crime refers to violent offenses known to the police, which includes murder, forcible rape, robbery, and aggravated assault. When states share the same rank, the next lower rank is omitted. Because of rounded data, states may have identical values shown but different ranks.

6. The following data represent the birth weights at an inner-city hospital in a large eastern city: 2.4, 3.3, 4.1, 5.0, 5.1, 5.2, 5.6, 5.8, 5.9, 5.9, 6.0, 6.1, 6.2, 6.3, 6.3, 6.4, 6.4, 6.5, 6.7, 6.8, 7.2, 7.4, 7.5, 7.5, 7.6, 7.6, 7.7, 7.8, 7.8, 7.9, 7.9, 8.3, 8.5, 8.8, 9.2, 9.7, 9.8, 9.9, 10.0, 10.3, 10.5 (a) (b) (c) (d) (e) (f) (g)

Plot this in a stem-and-leaf diagram. Find the sample mean x. Find the sample median. Find the sample standard deviation s. What proportion of the data lies within x ± 2s? Do the data appear to be approximately normal? If your answer to (f) is yes, what would you have estimated, based on your answers to (b) and (d), for (e)? *7. Let a and b be constants. Show that if yi = a + bxi for i = 1, . . . , n, then r, the sample correlation coefﬁcient of the data pairs xi , yi , i = 1, . . . , n, is given by (a) r = 1 when b > 0 (b) r = −1 when b < 0 (Hint: Use the deﬁnition of r, not its computational formula.) 8. The following data are taken from the book Researches on the Probability of Criminal and Civil Verdicts, published in 1837 by the French mathematician and probabilist Simeon Poisson. The book emphasized

Review Problems

legal applications of probability. The data refer to the number of people accused and convicted of crimes in France from 1825 to 1830.

Year

Number accused

Number convicted

1825

6652

4037

1826

6988

4348

1827

6929

4236

1828

7396

4551

1829

7373

4475

1830

6962

4130

(a) Determine the sample mean and sample median of the number accused. (b) Determine the sample mean and sample median of the number convicted. (c) Determine the sample standard deviation of the number accused. (d) Determine the sample standard deviation of the number convicted. (e) Would you expect the number accused and the number convicted to have a positive or a negative sample correlation coefﬁcient? (f) Determine the sample correlation coefﬁcient of the number of accused and number of convicted. (g) Determine the sample correlation coefﬁcient between the number accused and the percentage of these who are convicted. (h) Draw scatter diagrams for parts (f) and (g). (i) Guess at the value of the sample correlation coefﬁcient between the number of convicted and the percentage convicted. (j) Draw a scatter diagram for the variables in (i). (k) Determine the sample correlation coefﬁcient for the variables in part (i). 9. Recent studies have been inconclusive about the connection between coffee consumption and coronary heart disease. If a study indicated that consumers of large amounts of coffee appeared to have a greater chance of suffering heart attacks than did drinkers of moderate amounts or drinkers of no coffee at all, would this necessarily “prove” that excessive coffee drinking leads to an increased risk of heart attack? What other explanations are possible? 10. Recent studies have indicated that death rates for married middleaged people appear to be lower than for single middle-aged people. Does this mean that marriage tends to increase one’s life span? What other explanations are possible?

141

142

CH A P T E R 3: Using Statistics to Summarize Data Sets

11. A June 9, 1994, article in The New York Times noted a study showing that years with low inﬂation rates tend to be years with high average-productivity increases. The article claimed that this supported the Federal Reserve Board’s claim that a low rate of inﬂation tends to result in an increase in productivity. Do you think the study provides strong evidence for this claim? Explain your answer. 12. The following table gives the 2008 medicare enrollment as a percentage of the total population for each of the 50 states and the District of Columbia. (a) Find the sample mean of these data values. (b) Is your answer to part (a) necessarily equal to the percentage of the entire population that is enrolled in medicare? Why or why not? Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware District of Columbia Florida Georgia Hawaii Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi

17% 8% 13% 18% 12% 12% 15% 16% 13% 17% 12% 15% 14% 15% 15% 17% 15% 17% 15% 19% 13% 16% 15% 14% 16%

Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming

16% 16% 15% 13% 15% 15% 15% 15% 15% 16% 16% 16% 15% 18% 17% 16% 16% 16% 11% 10% 17% 14% 14% 20% 15% 14%

13. A sample of size n + m consists of numerical values from n men and m women. If x w is the sample mean of the women’s values, and x m is the sample mean of the men’s values what is the sample mean of the entire sample?

Review Problems

14. A random sample of individuals were rated as to their standing posture. In addition, the numbers of days of back pain each had experienced during the past year were also recorded. Surprisingly to the researcher these data indicated a positive correlation between good posture and number of days of back pain. Does this indicate that good posture causes back pain? 15. The following are the number of trafﬁc deaths in a sample of states, both for 2007 and 2008. Plot a scatter diagram and ﬁnd the sample correlation coefﬁcient for the data pairs. 2007 and 2008 Trafﬁc Fatalities per State State

2007

2008

WY IL MA NJ MD OR WA FL UT NH

149 1248 434 724 615 452 568 3221 291 129

159 1044 318 594 560 414 504 2986 271 139

143

This page intentionally left blank

CHAPTER 4

Probability Probability is the very guide of life. Cicero, De Natura

CONTENTS 4.1

Introduction ......................................................................... 146

4.2

Sample Space and Events of an Experiment ............................. 146 Problems ............................................................................. 150

4.3

Properties of Probability ........................................................ 153 Problems ............................................................................. 156

4.4

Experiments Having Equally Likely Outcomes ......................... 161 Problems ............................................................................. 164

4.5

Conditional Probability and Independence............................... 167 Problems ............................................................................. 177

*4.6 Bayes’ Theorem ................................................................... 185 Problems ............................................................................. 187 *4.7 Counting Principles .............................................................. 189 Problems ............................................................................. 195 Key Terms .................................................................................. 198 Summary .................................................................................... 200 Review Problems ......................................................................... 201 This chapter starts with consideration of an experiment whose outcome cannot be predicted with certainty. We deﬁne the events of this experiment. We then introduce the concept of the probability of an event, which is the probability that the outcome of the experiment is contained in the event. An interpretation of the Introductory Statistics, DOI: 10.1016/B978-0-12-374388-6.00004-1 © 2010, Elsevier Inc. All rights reserved.

145

146

CH A P T E R 4: Probability

probability of an event as being a long-term relative frequency is given. Properties of probabilities are discussed. The conditional probability of one event, given the occurrence of a second event, is introduced. We see what it means for events to be independent.

4.1 INTRODUCTION To gain information about the current leader in the next gubernatorial election, a representative sample of 100 voters has been polled. If 62 of those polled are in favor of the Republican candidate, can we conclude that a majority of the state’s voters favor this candidate? Or, is it possible that by chance the sample contained a much greater proportion of this candidate’s supporters than is contained in the general population and that the Democratic candidate is actually the current choice of a majority of the electorate? To answer these questions, it is necessary to know something about the chance that as many as 62 people in a representative sample of size 100 would favor a candidate who, in fact, is not favored by a majority of the entire population. Indeed, as a general rule, to be able to draw valid inferences about a population from a sample, one needs to know how likely it is that certain events will occur under various circumstances. The determination of the likelihood, or chance, that an event will occur is the subject matter of probability.

4.2 SAMPLE SPACE AND EVENTS OF AN EXPERIMENT The word probability is a commonly used term that relates to the chance that a particular event will occur when some experiment is performed, where we use the word experiment in a very broad sense. Indeed, an experiment for us is any process that produces an observation, or outcome. We are often concerned with an experiment whose outcome is not predictable, with certainty, in advance. Even though the outcome of the experiment will not be known in advance, we will suppose that the set of all possible outcomes is known. This set of all possible outcomes of the experiment is called the sample space and is denoted by S. Deﬁnition An experiment is any process that produces an observation or outcome. The set of all possible outcomes of an experiment is called the sample space. ■

Example 4.1 Some examples of experiments and their sample spaces are as follows. (a) If the outcome of the experiment is the gender of a child, then S = {g, b}

4.2 Sample Space and Events of an Experiment

where outcome g means that the child is a girl and b that it is a boy. (b) If the experiment consists of ﬂipping two coins and noting whether they land heads or tails, then S = {(H, H), (H, T), (T, H), (T, T)} The outcome is (H, H) if both coins land heads, (H, T) if the ﬁrst coin lands heads and the second tails, (T, H) if the ﬁrst is tails and the second is heads, and (T, T) if both coins land tails. (c) If the outcome of the experiment is the order of ﬁnish in a race among 7 horses having positions 1, 2, 3, 4, 5, 6, 7, then S = {all orderings of 1, 2, 3, 4, 5, 6, 7} The outcome (4, 1, 6, 7, 5, 3, 2) means, for instance, that the number 4 horse comes in ﬁrst, the number 1 horse comes in second, and so on. (d) Consider an experiment that consists of rolling two six-sided dice and noting the sides facing up. Calling one of the dice die 1 and the other die 2, we can represent the outcome of this experiment by the pair of upturned values on these dice. If we let (i, j) denote the outcome in which die 1 has value i and die 2 has value j, then the sample space of this experiment is S = {(1, 1), (1, 2), (1, 3), (1, 4), (1, 5), (1, 6), (2, 1), (2, 2), (2, 3), (2, 4), (2, 5), (2, 6), (3, 1), (3, 2), (3, 3), (3, 4), (3, 5), (3, 6), (4, 1), (4, 2), (4, 3), (4, 4), (4, 5), (4, 6), (5, 1), (5, 2), (5, 3), (5, 4), (5, 5), (5, 6), (6, 1), (6, 2), (6, 3), (6, 4), (6, 5), (6, 6)} ■ Any set of outcomes of the experiment is called an event. That is, an event is a subset of the sample space. Events will be denoted by the capital letters A, B, C, and so on. ■

Example 4.2 In Example 4.1(a), if A = {g}, then A is the event that the child is a girl. Similarly, if B = {b}, then B is the event that the child is a boy. In Example 4.1(b), if A = {(H, H), (H, T)}, then A is the event that the ﬁrst coin lands on heads. In Example 4.1(c), if A = {all outcomes in S starting with 2} then A is the event that horse number 2 wins the race.

147

148

CH A P T E R 4: Probability

In Example 4.1(d), if A = {(1, 6), (2, 5), (3, 4), (4, 3), (5, 2), (6, 1)} then A is the event that the sum of the dice is 7.

■

Deﬁnition Any set of outcomes of the experiment is called an event. We designate events by the letters A, B, C, and so on. We say that the event A occurs whenever the outcome is contained in A. For any two events A and B, we deﬁne the new event A ∪ B, called the union of events A and B, to consist of all outcomes that are in A or in B or in both A and B. That is, the event A ∪ B will occur if either A or B occurs. In Example 4.1(a), if A = {g} is the event that the child is a girl and B = {b} is the event that it is a boy, then A ∪ B = {g, b}. That is, A ∪ B is the whole sample space S. In Example 4.1(c), let A = {all outcomes starting with 4} be the event that the number 4 horse wins; and let B = {all outcomes whose second element is 2} be the event that the number 2 horse comes in second. Then A ∪ B is the event that either the number 4 horse wins or the number 2 horse comes in second or both. A graphical representation of events that is very useful is the Venn diagram. The sample space S is represented as consisting of all the points in a large rectangle, and events are represented as consisting of all the points in circles within the rectangle. Events of interest are indicated by shading appropriate regions of the diagram. The colored region of Fig. 4.1 represents the union of events A and B. For any two events A and B, we deﬁne the intersection of A and B to consist of all outcomes that are both in A and in B. That is, the intersection will occur if both A and B occur. We denote the intersection of A and B by A ∩ B. The colored region of Fig. 4.2 represents the intersection of events A and B. In Example 4.1(b), if A = {(H, H), (H, T)} is the event that the ﬁrst coin lands heads and B = {(H, T), (T, T)} is the event that the second coin lands tails, then A ∩ B = {(H, T)} is the event that the ﬁrst coin lands heads and the second lands tails. In Example 4.1(c), if A is the event that the number 2 horse wins and B is the event that the number 3 horse wins, then the event A ∩ B does not contain any outcomes and so cannot occur. We call the event without any outcomes the null event, and

4.2 Sample Space and Events of an Experiment

FIGURE 4.1 A Venn diagram: shaded region is A ∪ B.

FIGURE 4.2 Shaded region is A ∩ B.

FIGURE 4.3 A and B are disjoint events.

designate it as Ø. If the intersection of A and B is the null event, then since A and B cannot simultaneously occur, we say that A and B are disjoint, or mutually exclusive. Two disjoint events are pictured in the Venn diagram of Fig. 4.3. For any event A we deﬁne the event Ac , called the complement of A, to consist of all outcomes in the sample space that are not in A. That is, Ac will occur when A does not, and vice versa. For instance, in Example 4.1(a), if A = {g} is the event that the child is a girl, then Ac = {b} is the event that it is a boy. Also note that the complement of the sample space is the null set, that is, Sc = Ø. Figure 4.4 indicates Ac , the complement of event A. We can also deﬁne unions and intersections of more than two events. For instance, the union of events A, B, and C, written A ∪ B ∪ C, consists of all the outcomes

149

150

CH A P T E R 4: Probability

FIGURE 4.4 Shaded region is Ac .

of the experiment that are in A or in B or in C. Thus, A ∪ B ∪ C will occur if at least one of these events occurs. Similarly, the intersection A ∩ B ∩ C consists of the outcomes that are in all the events A, B, and C. Thus, the intersection will occur only if all the events occur. We say that events A, B, and C are disjoint if no two of them can simultaneously occur.

PROBLEMS 1. A box contains three balls—one red, one blue, and one yellow. Consider an experiment that consists of withdrawing a ball from the box, replacing it, and withdrawing a second ball. (a) What is the sample space of this experiment? (b) What is the event that the ﬁrst ball drawn is yellow? (c) What is the event that the same ball is drawn twice? 2. Repeat Prob. 1 when the second ball is drawn without replacement of the ﬁrst ball. 3. Audrey and her boyfriend Charles must both choose which colleges they will attend in the coming fall. Audrey was accepted at the University of Michigan (MI), Reed College (OR), San Jose State College (CA), Yale University (CT), and Oregon State University (OR). Charles was accepted at Oregon State University and San Jose State College. Let the outcome of the experiment consist of the colleges that Audrey and Charles choose to attend. (a) List all the outcomes in sample space S. (b) List all the outcomes in the event that Audrey and Charles attend the same school. (c) List all the outcomes in the event that Audrey and Charles attend different schools.

4.2 Sample Space and Events of an Experiment

4.

5.

6.

7.

8.

(d) List all the outcomes in the event that Audrey and Charles attend schools in the same state. An experiment consists of ﬂipping a coin three times and each time noting whether it lands heads or tails. (a) What is the sample space of this experiment? (b) What is the event that tails occur more often than heads? Family members have decided that their next vacation will be either in France or in Canada. If they go to France, they can either ﬂy or take a boat. If they go to Canada, they can drive, take a train, or ﬂy. Letting the outcome of the experiment be the location of their vacation and their mode of travel, list all the points in sample space S. Also list all the outcomes in A, where A is the event that the family ﬂies to the destination. The New York Yankees and the Chicago White Sox are playing three games this weekend. Assuming that all games are played to a conclusion and that we are interested only in which team wins each game, list all the outcomes in sample space S. Also list all the outcomes in A, where A is the event that the Yankees win more games than the White Sox. Let S = {1, 2, 3, 4, 5, 6}, A = {1, 3, 5}, B = {4, 6}, and C = {1, 4}. Find (a) A ∩ B (b) B ∪ C (c) A ∪ (B ∩ C) (d) (A ∪ B)c Note: The operations within parentheses are performed ﬁrst. For instance, in (c) ﬁrst determine the intersection of B and C, and then take the union of A and that set. A cafeteria offers a three-course meal. One chooses a main course, a starch, and a dessert. The possible choices are as follows: Meal Main course Starch course Dessert

Choices Chicken or roast beef Pasta or rice or potatoes Ice cream or gelatin or apple pie

An individual is to choose one course from each category. (a) List all the outcomes in the sample space. (b) Let A be the event that ice cream is chosen. List all the outcomes in A. (c) Let B be the event that chicken is chosen. List all the outcomes in B.

151

152

CH A P T E R 4: Probability

9.

10.

11.

12.

(d) List all the outcomes in the event A ∩ B. (e) Let C be the event that rice is chosen. List all the outcomes in C. (f) List all the outcomes in the event A ∩ B ∩ C. A hospital administrator codes patients according to whether they have insurance and according to their condition, which is rated as good, fair, serious, or critical. The administrator records a 0 if a patient has no insurance and a 1 if he or she does, and then records one of the letters g, f , s, or c, depending on the patient’s condition. Thus, for instance, the coding 1, g is used for a patient with insurance who is in good condition. Consider an experiment that consists of the coding of a new patient. (a) List the sample space of this experiment. (b) Specify the event corresponding to the patient’s being in serious or critical condition and having no medical insurance. (c) Specify the event corresponding to the patient’s being in either good or fair condition. (d) Specify the event corresponding to the patient’s having insurance. The following pairs of events E and F relate to the same experiment. Tell in each case whether E and F are disjoint events. (a) A die is rolled. Event E is that it lands on an even number, and F is the event that it lands on an odd number. (b) A die is rolled. Event E is that it lands on 3, and F is the event that it lands on an even number. (c) A person is chosen. Event E is that this person was born in the United States, and F is the event that this person is a U.S. citizen. (d) A man is chosen. Event E is that he is over 30 years of age, and F is the event that he has been married for over 30 years. (e) A woman waiting in line to register her car at the department of motor vehicles is chosen. Event E is that the car is made in the United States, and F is the event that it is made in a foreign country. Let A be the event that a rolled die lands on an even number. (a) Describe in words the event Ac . (b) Describe in words the event (Ac )c . (c) In general, let A be an event. What is the complement of its complement? That is, what is (Ac )c ? Two dice are rolled. Let A be the event that the sum of the dice is even, let B be the event that the ﬁrst die lands on 1, and let C be the event that the sum of the dice is 6. Describe the following events. (a) A ∩ B (b) A ∪ B (c) B ∩ C (d) Bc

4.3 Properties of Probability

(e) Ac ∩ C (f) A ∩ B ∩ C 13. Let A, B, and C be events. Use Venn diagrams to represent the event that of A, B, and C (a) Only A occurs. (b) Both A and B occur, but C does not. (c) At least one event occurs. (d) At least two of the events occur. (e) All three events occur.

4.3 PROPERTIES OF PROBABILITY It is an empirical fact that if an experiment is continually repeated under the same conditions, then, for any event A, the proportion of times that the outcome is contained in A approaches some value as the number of repetitions increases. For example, if a coin is continually ﬂipped, then the proportion of ﬂips landing on tails will approach some value as the number of ﬂips increases. It is this long-run proportion, or relative frequency, that we often have in mind when we speak of the probability of an event. Consider an experiment whose sample space is S. We suppose that for each event A there is a number, denoted P(A) and called the probability of event A, that is in accord with the following three properties. PROPERTY 1: For any event A, the probability of A is a number between 0 and 1. That is, 0 ≤ P(A) ≤ 1 PROPERTY 2: The probability of sample space S is 1. Symbolically, P(S) = 1 PROPERTY 3: The probability of the union of disjoint events is equal to the sum of the probabilities of these events. For instance, if A and B are disjoint, then P(A ∪ B) = P(A) + P(B) The quantity P(A) represents the probability that the outcome of the experiment is contained in event A. Property 1 states that the probability that the outcome of the experiment is contained in A is some value between 0 and 1. Property 2 states that, with probability 1, the outcome of the experiment will be an element of sample space S. Property 3 states that if events A and B cannot simultaneously occur, then the probability that the outcome of the experiment is contained in either A or B is equal to the sum of the probability that it is in A and the probability that it is in B.

153

154

CH A P T E R 4: Probability

If we interpret P(A) as the long-run relative frequency of event A, then the stated conditions are satisﬁed. The proportion of experiments in which the outcome is contained in A would certainly be a number between 0 and 1. The proportion of experiments in which the outcome is contained in S is 1 since all outcomes are contained in sample space S. Finally, if A and B have no outcomes in common, then the proportion of experiments whose outcome is in either A or B is equal to the proportion whose outcome is in A plus the proportion whose outcome is in B. For instance, if the proportion of time that a pair of rolled dice sums to 7 is 1/6 and the proportion of time that they sum to 11 is 1/18, then the proportion of time that they sum to either 7 or 11 is 1/6 + 1/18 = 2/9. Properties 1, 2, and 3 can be used to establish some general results concerning probabilities. For instance, since A and Ac are disjoint events whose union is the entire sample space, we can write S = A ∪ Ac Using properties 2 and 3 now yields the following. 1 = P(S)

by property 2 c

= P(A ∪ A ) = P(A) + P(Ac )

by property 3

Therefore, we see that P(Ac ) = 1 − P(A) In words, the probability that the outcome of the experiment is not contained in A is 1 minus the probability that it is. For instance, if the probability of obtaining heads on the toss of a coin is 0.4, then the probability of obtaining tails is 0.6. The following formula relates the probability of the union of events A and B, which are not necessarily disjoint, to P(A), P(B), and the probability of the intersection of A and B. It is often called the addition rule of probability. Addition Rule For any events A and B, P(A ∪ B) = P(A) + P(B) − P(A ∩ B) To see why the addition rule holds, note that P(A ∪ B) is the probability of all outcomes that are either in A or in B. On the other hand, P(A) + P(B) is the probability of all the outcomes that are in A plus the probability of all the outcomes that are in B. Since any outcome that is in both A and B is counted twice in P(A) + P(B)

4.3 Properties of Probability

FIGURE 4.5 P(A ∪ B) = P(A) + P(B) − P(A ∩ B).

and only once in P(A ∪ B) (see Fig. 4.5), it follows that P(A) + P(B) = P(A ∪ B) + P(A ∩ B) Subtracting P(A ∩ B) from both sides of the preceding equation gives the addition rule. Example 4.3 illustrates the use of the addition rule. ■

Example 4.3 A certain retail establishment accepts either the American Express or the VISA credit card. A total of 22 percent of its customers carry an American Express card, 58 percent carry a VISA credit card, and 14 percent carry both. What is the probability that a customer will have at least one of these cards? Solution Let A denote the event that the customer has an American Express card, and let B be the event that she or he has a VISA card. The given information yields P(A) = 0.22

P(B) = 0.58

P(A ∩ B) = 0.14

By the additive rule, the desired probability P(A ∪ B) is P(A ∪ B) = 0.22 + 0.58 − 0.14 = 0.66 That is, 66 percent of the establishment’s customers carry at least one of the cards that it will accept. ■ As an illustration of the interpretation of probability as a long-run relative frequency, we have simulated 10,000 ﬂips of a perfectly symmetric coin. The total numbers of heads and tails that occurred in the ﬁrst 10, 50, 100, 500, 2000, 6000, 8000, and 10,000 ﬂips, along with the proportion of them that was heads, are

155

156

CH A P T E R 4: Probability

presented in Table 4.1. Note how the proportion of the ﬂips that lands heads becomes very close to 0.5 as the number of ﬂips increases. Table 4.1 10,000 Flips of a Symmetric Coin n

Number of heads in ﬁrst n ﬂips

Number of tails in ﬁrst n ﬂips

Proportion of ﬁrst n ﬂips that lands on heads

3 21 46 248 1,004 3,011 3,974 5,011

7 29 54 252 996 2,989 4,026 4,989

0.3 0.42 0.46 0.496 0.502 0.5018 0.4968 0.5011

10 50 100 500 2,000 6,000 8,000 10,000

Table 4.2 10,000 Rolls of a Symmetric Die i

Frequency of outcome Relative frequency

1

2

3

4

5

6

1724 0.1724

1664 0.1664

1628 0.1628

1648 0.1648

1672 0.1672

1664 0.1664

Note: 1/6 = 0.166667.

The results of 10,000 simulated rolls of a perfectly symmetric die are presented in Table 4.2.

PROBLEMS 1. Suppose the sample space of an experiment is S = {1, 2, 3, 4, 5, 6} Let A i denote the event consisting of the single outcome i, and suppose that P(A1 ) = 0.1

P(A4 ) = 0.15

P(A2 ) = 0.2

P(A5 ) = 0.1

P(A3 ) = 0.15

P(A6 ) = 0.3

That is, the outcome of the experiment is 1 with probability 0.1, it is 2 with probability 0.2, it is 3 with probability 0.15, and so on. Let events E, F, and G be as follows: E = {1, 3, 5}

F = {2, 4, 6}

G = {1, 4, 6}

4.3 Properties of Probability

A few years later a similar story took place in France, where a gambler known as Chevalier de Mere resided. De Mere, a strong amateur mathematician as well as a gambler, had an acquaintance with the brilliant mathematician Blaise Pascal. It was to Pascal that de Mere turned for help in his more difﬁcult gaming questions. One particular problem, known as the problem of the points, concerned the equitable division of stakes when two players are interrupted in the midst of a game of chance. Pascal found this problem particularly intriguing and, in 1654, wrote to the mathematician Pierre Fermat about it. Their resulting exchange of letters not only led to a solution of this problem but also laid the framework for the solution of many other problems connected with games of chance. Their celebrated correspondence, cited by some as the birth date of probability, stimulated interest in probability among some of the foremost European mathematicians of the time. For instance, the young Dutch genius Ludwig Huyghens came to Paris to discuss the new subject, and activity in this new ﬁeld grew rapidly.

Find (a) P(E), P(F), P(G) (b) P(E ∪ F) (c) P(E ∪ G) (d) P(F ∪ G) (e) P(E ∪ F ∪ G) (f) P(E ∩ F) (g) P(F ∩ G) (h) P(E ∩ G) (i) P(E ∩ F ∩ G) 2. If A and B are disjoint events for which P(A) = 0.2 and P(B) = 0.5, ﬁnd (a) P(Ac ) (b) P(A ∪ B) (c) P(A ∩ B) (d) P(Ac ∩ B) 3. Phenylketonuria is a genetic disorder that produces mental retardation. About one child in every 10,000 live births in the United States has phenylketonuria. What is the probability that the next child born in a Houston hospital has phenylketonuria?

Pierre Fermat

(Bettmann)

The notion that chance, or probability, can be treated numerically is relatively recent. Indeed, for most of recorded history it was felt that what occurred in life was determined by forces that were beyond one’s ability to understand. It was only during the ﬁrst half of the 17th century, near the end of the Renaissance, that people became curious about the world and the laws governing its operation. Among the curious were the gamblers. A group of Italian gamblers, unable to answer certain questions concerning dice, approached the famous scientist Galileo. Galileo, though busy with other work, found their problems to be of interest and not only provided solutions but also wrote a short treatise on games of chance.

(Bettmann)

Historical Perspective

Blaise Pascal

157

158

CH A P T E R 4: Probability

4. A certain person encounters three trafﬁc lights when driving to work. Suppose that the following represent the probabilities of the total number of red lights that she has to stop for: P(0 red lights) = 0.14 P(1 red light) = 0.36 P(2 red lights) = 0.34 P(3 red lights) = 0.16 (a) What is the probability that she stops for at least one red light when driving to work? (b) What is the probability that she stops for more than two red lights? 5. If A and B are disjoint events, is the following possible? P(A) + P(B) = 1.2 What if A and B are not disjoint? 6. If the probability of drawing a king from a deck of pinochle cards is 1/6 and the probability of drawing an ace is 1/6, what is the probability of drawing either an ace or a king? 7. Suppose that the demand for Christmas trees from a certain dealer will be 1100 1400 1600 2000

with probability 0.2 with probability 0.3 with probability 0.4 with probability 0.1

Find the probability that the dealer will be able to sell his entire stock if he purchases (a) 1100 trees (b) 1400 trees (c) 1600 trees (d) 2000 trees 8. The Japanese automobile company Lexus has established a reputation for quality control. Recent statistics indicate that a newly purchased Lexus ES 350 will have 0 defects 1 defect 2 defects 3 defects 4 defects 5 or more defects

with probability 0.12 with probability 0.18 with probability 0.25 with probability 0.20 with probability 0.15 with probability 0.10

4.3 Properties of Probability

If you purchase a new Lexus ES 350, ﬁnd the probability that it will have (a) 2 or fewer defects (b) 4 or more defects (c) Between (inclusive) 1 and 3 defects Let p denote the probability it will have an even number of defects. Whereas the information given above does not enable us to specify the value of p, ﬁnd the (d) Largest (e) Smallest value of p which is consistent with the preceding. 9. When typing a ﬁve-page manuscript, a certain typist makes 0 errors 1 error 2 errors 3 errors 4 or more errors

with probability 0.20 with probability 0.35 with probability 0.25 with probability 0.15 with probability 0.05

If you give such a manuscript to this typist, ﬁnd the probability that it will contain (a) 3 or fewer errors (b) 2 or fewer errors (c) 0 errors 10. The following table is a modern version of a life table, which was ﬁrst developed by John Graunt in 1662. It gives the probabilities that a newly born member of a certain speciﬁed group will die in his or her ith decade of life, for i ranging from 1 to 10. The ﬁrst decade starts with birth and ends with an individual’s 10th birthday. The second decade starts at age 10 and ends at the 20th birthday, and so on. Life Table

Decade 1 2 3 4 5

Probability of death 0.062 0.012 0.024 0.033 0.063

Decade 6 7 8 9 10

Probability of death 0.124 0.215 0.271 0.168 0.028

For example, the probability that a newborn child dies in her or his ﬁfties is 0.124. Find the probability that a newborn will (a) Die between the ages of 30 and 60 (b) Not survive to age 40 (c) Survive to age 80

159

160

CH A P T E R 4: Probability

11. The family picnic scheduled for tomorrow will be postponed if it is either cloudy or rainy. The weather report states that there is a 40 percent chance of rain tomorrow, a 50 percent chance of cloudiness, and a 20 percent chance that it will be both cloudy and rainy. What is the probability that the picnic will be postponed? 12. In Example 4.3, what proportion of customers has neither an American Express nor a VISA card? 13. It is estimated that 30 percent of all adults in the United States are obese and that 3 percent suffer from diabetes. If 2 percent of the population both is obese and suffers from diabetes, what percentage of the population either is obese or suffers from diabetes? 14. Welds of tubular joints can have two types of defects, which we call A and B. Each weld produced has defect A with probability 0.064, defect B with probability 0.043, and both defects with probability 0.025. Find the proportion of welds that has (a) Either defect A or defect B (b) Neither defect 15. A customer that goes to the suit department of a certain store will purchase a suit with probability 0.3. The customer will purchase a tie with probability 0.2 and will purchase both a suit and a tie with probability 0.1. What proportion of customers purchases neither a suit nor a tie? 16. Anita has a 40 percent chance of receiving an A grade in statistics, a 60 percent chance of receiving an A in physics, and an 86 percent chance of receiving an A in either statistics or physics. Find the probability that she (a) Does not receive an A in either statistics or physics (b) Receives A’s in both statistics and physics 17. This problem uses a Venn diagram to present a formal proof of the addition rule. Events A and B are represented by circles in the Venn diagram. In terms of A and B, describe the region labeled (a) I (b) II (c) III

4.4 Experiments Having Equally Likely Outcomes

Express, in terms of P(I), P(II), and P(III), (d) P(A ∪ B) (e) P(A) (f) P(B) (g) P(A ∩ B) (h) Conclude that P(A ∪ B) = P(A) + P(B) − P(A ∩ B)

4.4 EXPERIMENTS HAVING EQUALLY LIKELY OUTCOMES For certain experiments it is natural to assume that each outcome in the sample space S is equally likely to occur. That is, if sample space S consists of N outcomes, say, S = {1, 2, . . . , N}, then it is often reasonable to suppose that P({1}) = P({2}) = · · · = P({N}) In this expression, P({i}) is the probability of the event consisting of the single outcome i; that is, it is the probability that the outcome of the experiment is i. Using the properties of probability, we can show that the foregoing implies that the probability of any event A is equal to the proportion of the outcomes in the sample space that is in A. That is, P(A) =

■

number of outcomes in S that are in A N

Example 4.4 In a survey of 420 members of a retirement center, it was found that 144 are smokers and 276 are not. If a member is selected in such a way that each of the members is equally likely to be the one selected, what is the probability that person is a smoker? Solution There are 420 outcomes in the sample space of the experiment of selecting a member of the center. Namely, the outcome is the person selected. Since there are 144 outcomes in the event that the selected person is a smoker, it follows that the probability of this event is P{smoker} =

144 12 = 420 35

■

161

162

CH A P T E R 4: Probability

■

Example 4.5 Suppose that when two dice are rolled, each of the 36 possible outcomes given in Example 4.1(d) is equally likely. Find the probability that the sum of the dice is 6 and that it is 7. Solution If we let A denote the event that the sum of the dice is 6 and B that it is 7, then A = {(1, 5), (2, 4), (3, 3), (4, 2), (5, 1)} and B = {(1, 6), (2, 5), (3, 4), (4, 3), (5, 2), (6, 1)} Therefore, since A contains 5 outcomes and B contains 6, we see that P(A) = P{sum is 6} = 5/36 P(B) = P{sum is 7} = 6/36 = 1/6

■

■

Example 4.6 One man and one woman are to be selected from a group that consists of 10 married couples. If all possible selections are equally likely, what is the probability that the woman and man selected are married to each other? Solution Once the man is selected, there are 10 possible choices of the woman. Since one of these 10 choices is the wife of the man chosen, we see that the desired probability is 1/10. ■

When each outcome of the sample space is equally likely to be the outcome of the experiment, we say that an element of the sample space is randomly selected.

■

Example 4.7 An elementary school is offering two optional language classes, one in French and the other in Spanish. These classes are open to any of the 120 upper-grade students in the school. Suppose there are 32 students in the French class, 36 in the Spanish class, and a total of 8 who are in both classes. If an upper-grade student is randomly chosen, what is the probability that this student is enrolled in at least one of these classes?

4.4 Experiments Having Equally Likely Outcomes

Solution Let A and B denote, respectively, the events that the randomly chosen student is enrolled in the French class and is enrolled in the Spanish class. We will determine P(A ∪ B), the probability that the student is enrolled in either French or Spanish, by using the addition rule P(A ∪ B) = P(A) + P(B) − P(A ∩ B) Since 32 of the 120 students are enrolled in the French class, 36 of the 120 are in the Spanish class, and 8 of the 120 are in both classes, we have P(A) =

32 , 120

P(B) =

36 , 120

and

P(A ∩ B) =

8 120

Therefore, P(A ∪ B) =

32 36 8 1 + − = 120 120 120 2

That is, the probability that a randomly chosen student is taking at least one of the language classes is 1/2. ■ ■

Example 4.8 Table 4.3 lists the earnings frequencies of all full-time workers who are at least 15 years old, classiﬁed according to their annual salary and gender.

Table 4.3 Earnings of Workers by Sex, 1989 Earnings group (in $1000) 100 Total

Number

Distribution (percent)

Women

Men

Women

Men

427,000 440,000 1,274,000 1,982,000 6,291,000 6,555,000 5,169,000 8,255,000 947,000

548,000 358,000 889,000 1,454,000 5,081,000 6,386,000 6,648,000 20,984,000 7,377,000

1.4 1.4 4.1 6.3 20.1 20.9 16.5 26.3 3.0

1.1 .7 1.8 2.9 10.2 12.9 13.4 42.1 14.9

31,340,000

49,678,000

100.0

100.0

Source: Department of Commerce, Bureau of the Census.

163

164

CH A P T E R 4: Probability

Suppose one of these workers is randomly chosen. Find the probability that this person is (a) A woman (b) A man (c) A man earning under $30,000 (d) A woman earning over $50,000 Solution (a) Since 31,340,000 of the 31,340,000 + 49,678,000 = 81,018,000 workers are women, it follows that the probability that a randomly chosen worker is a woman is 31,340,000 ≈ .3868 81,018,000 That is, there is approximately a 38.7 percent chance that the randomly selected worker is a woman. (b) Since the event that the randomly selected worker is a man is the complement of the event that the worker is a woman, we see from (a) that the probability is approximately 1 − 0.3868 = 0.6132. (c) Since (in thousands) the number of men earning under $30,000 is 548 + 358 + 889 + 1454 + 5081 = 8330 we see that the desired probability is 8330/81,018 ≈ .1028. That is, there is approximately a 10.3 percent chance that the person selected is a man with an income under $30,000. (d) The probability that the person selected is a woman with an income above $50,000 is 8255 + 947 ≈ .1136 81,018 That is, there is approximately an 11.4 percent chance that the person selected is a woman with an income above $50,000. ■

PROBLEMS 1. In an experiment involving smoke detectors, an alarm was set off at a college dormitory at 3 a.m. Out of 216 residents of the dormitory, 128 slept through the alarm. If one of the residents is randomly chosen, what is the probability that this person did not sleep through the alarm? 2. Among 32 dieters following a similar routine, 18 lost weight, 5 gained weight, and 9 remained the same weight. If one of these dieters is randomly chosen, ﬁnd the probability that he or she

4.4 Experiments Having Equally Likely Outcomes

(a) Gained weight (b) Lost weight (c) Neither lost nor gained weight 3. One card is to be selected at random from an ordinary deck of 52 cards. Find the probability that the selected card is (a) An ace (b) Not an ace (c) A spade (d) The ace of spades 4. The following table lists the 10 countries with the highest production of meat.

Country China United States Russia Germany France Brazil Argentina Britain Italy Australia

Meat production (thousands of metric tons) 20,136 17,564 12,698 6,395 3,853 3,003 2,951 2,440 2,413 2,373

Suppose a World Health Organization committee is formed to discuss the long-term ramiﬁcations of producing such quantities of meat. Suppose further that it consists of one representative from each of these countries. If the chair of this committee is then randomly chosen, ﬁnd the probability that this person will be from a country whose production of meat (in thousands of metric tons) (a) Exceeds 10,000 (b) Is under 3500 (c) Is between 4000 and 6000 (d) Is less than 2000 5. Suppose that distinct integer values are written on each of 3 cards. These cards are then randomly given the designations A, B, and C. The values on cards A and B are then compared. If the smaller of these values is then compared with the value on card C, what is the probability that it is also smaller than the value on card C ? 6. A bag containing pennies and dimes has 4 times as many dimes as pennies. One coin is drawn. Assuming that the drawn coin is equally likely to be any of the coins, what is the probability that it is a dime? 7. A total of 44 out of 100 patients at a rehabilitation center are signed up for a special exercise program that consists of a swimming class and a

165

166

CH A P T E R 4: Probability

8.

9.

10.

11. 12.

13.

14.

15.

calisthenics class. Each of these 44 patients takes at least one of these classes. Suppose that there are 26 patients in the swimming class and 28 in the calisthenics class. Find the probability that a randomly chosen patient at the center is (a) Not in the exercise program (b) Enrolled in both classes Of the families in a certain community, 20 percent have a cat, 32 percent have a dog, and 12 percent have both a cat and a dog. (a) If a family is chosen at random, what is the probability it has neither a dog nor a cat? (b) If the community consists of 1000 families, how many of them have either a cat or a dog? Of the students at a girls’ school, 60 percent wear neither a ring nor a necklace, 20 percent wear a ring, and 30 percent wear a necklace. If one of them is randomly chosen, ﬁnd the probability that she is wearing (a) A ring or a necklace (b) A ring and a necklace A sports club has 120 members, of whom 44 play tennis, 30 play squash, and 18 play both tennis and squash. If a member is chosen at random, ﬁnd the probability that this person (a) Does not play tennis (b) Does not play squash (c) Plays neither tennis nor squash In Prob. 10, how many members play either tennis or squash? If two dice are rolled, ﬁnd the probability that the sum of the dice is (a) Either 7 or 11 (b) One of the values 2, 3, or 12 (c) An even number Suppose 2 people are randomly chosen from a set of 20 people that consists of 10 married couples. What is the probability that the 2 people are married to each other? (Hint: After the initial person is chosen, the next one is equally likely to be any of the remaining people.) Find the probability that a randomly chosen worker in Example 4.8 (a) Earns under $15,000 (b) Is a woman who earns between $20,000 and $40,000 (c) Earns under $50,000 A real estate agent has a set of 10 keys, one of which will open the front door of a house he is trying to show to a client. If the keys are tried in a completely random order, ﬁnd the probability that (a) The ﬁrst key opens the door (b) All 10 keys are tried

4.5 Conditional Probability and Independence

16. A group of 5 girls and 4 boys is randomly lined up. (a) What is the probability that the person in the second position is a boy? (b) What is the probability that Charles (one of the boys) is in the second position? 17. The following data are from the U.S. National Oceanic and Atmospheric Administration. They give the average number of days in each month with precipitation of 0.01 inch or more for Washington, D.C. Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec. 10

9

11

10

11

10

10

9

8

7

8

9

Find the probability you will encounter rain if you are planning to visit Washington, D.C., next (a) January 5 (b) August 12 (c) April 15 (d) May 15 (e) October 12

4.5 CONDITIONAL PROBABILITY AND INDEPENDENCE We are often interested in determining probabilities when some partial information concerning the outcome of the experiment is available. In such situations, the probabilities are called conditional probabilities. As an example of a conditional probability, suppose two dice are to be rolled. Then, as noted in Example 4.1(d), the sample space of this experiment is the set of 36 outcomes (i, j), where both i and j range from 1 through 6. The outcome (i, j) results when the ﬁrst die lands on i and the second on j. Suppose that each of the 36 possible outcomes is equally likely to occur and thus has probability 1/36. (When this is the case, we say that the dice are fair.) Suppose further that the ﬁrst die lands on 4. Given this information, what is the resulting probability that the sum of the dice is 10? To determine this probability, we reason as follows. Given that the ﬁrst die lands on 4, there are 6 possible outcomes of the experiment, namely, (4, 1), (4, 2), (4, 3), (4, 4), (4, 5), (4, 6) In addition, since these outcomes initially had the same probabilities of occurrence, they should still have equal probabilities. That is, given that the ﬁrst die

167

168

CH A P T E R 4: Probability

lands on 4, the conditional probability of each of the outcomes should be 1/6. Since in only one of the outcomes is the sum of the dice equal to 10, namely, the outcome (4, 6), it follows that the conditional probability that the sum is 10, given that the ﬁrst die lands on 4, is 1/6. If we let B denote the event that the sum of the dice is 10 and let A denote the event that the ﬁrst die lands on 4, then the probability obtained is called the conditional probability of B given that A has occurred. It is denoted by P(B|A) A general formula for P(B|A) can be derived by an argument similar to the one used earlier. Suppose that the outcome of the experiment is contained in A. Now, in order for the outcome also to be in B, it must be in both A and B; that is, it must be in A ∩ B. However, since we know that the outcome is in A, it follows that A becomes our new (or reduced) sample space, and the probability that event A ∩ B occurs is the probability of A ∩ B relative to the probability of A. That is (see Fig. 4.6), P(A ∩ B) P(B|A) = P(A) This deﬁnition of conditional probability is consistent with the interpretation of probability as being a long-run relative frequency. To show this, suppose that a large number, call it n, of repetitions of the experiment are performed. We will now argue that if we consider only those experiments in which A occurs, then P(B|A) will equal the long-run proportion of them in which B also occurs. To see this, note that since P(A) is the long-run proportion of experiments in which A occurs, it follows that in n repetitions of the experiment, A will occur approximately nP(A) times. Similarly, in approximately nP(A ∩ B) of these experiments, both A and B will occur. Hence, out of the approximately nP(A) experiments for which the outcome is contained in A, approximately nP(A ∩ B) of them will also have their outcomes in B. Therefore, of those experiments whose outcomes are in A, the proportion whose outcome is also in B is approximately equal to P(A ∩ B) nP(A ∩ B) = nP(A) P(A)

FIGURE 4.6 P(B|A) = P(A∩B) P(A) .

4.5 Conditional Probability and Independence

Since this approximation becomes exact as n becomes larger and larger, we see that we have given the appropriate deﬁnition of the conditional probability of B given that A has occurred.

■

Example 4.9 As a further check of the preceding formula for the conditional probability, use it to compute the conditional probability that the sum of a pair of rolled dice is 10, given that the ﬁrst die lands on 4. Solution Letting B denote the event that the sum of the dice is 10 and A the event that the ﬁrst die lands on 4, we have P(B|A) =

P(A ∩ B) P(A)

=

P({(4, 6)}) P({(4, 1), (4, 2), (4, 3), (4, 4), (4, 5), (4, 6)})

=

1/36 1 = 6/36 6

Therefore, we obtain the same result as before.

■

■

Example 4.10 The organization that employs Jacobi is organizing a parent-daughter dinner for those employees having at least one daughter. Each of these employees is asked to attend along with one of his or her daughters. If Jacobi is known to have two children, what is the conditional probability that they are both girls given that Jacobi is invited to the dinner? Assume the sample space S is given by S = {(g, g), (g, b), (b, g), (b, b)} and that all these outcomes are equally likely, where the outcome (g, b) means, for instance, that Jacobi’s oldest child is a girl and youngest is a boy. Solution Since Jacobi is invited to the dinner, we know that at least one of Jacobi’s children is a girl. Letting B denote the event that both of them are girls and A the event that at least one is a girl, we see that the desired probability is P(B|A). This

169

170

CH A P T E R 4: Probability

is determined as follows: P(A ∩ B) P(A) P({g, g}) = P({(g, g), (g, b), (b, g)}) 1 1/4 = = 3/4 3

P(B|A) =

That is, the conditional probability that both of Jacobi’s children are girls given that at least one is a girl is 1/3. Many students incorrectly suppose that this conditional probability is 1/2, reasoning that the Jacobi child not attending the dinner is equally likely to be a boy or a girl. Their mistake lies in assuming that these two possibilities are equally likely, for initially there were 4 equally likely outcomes. The information that at least one of the children is a girl is equivalent to knowing that the outcome is not (b, b). Thus we are left with the 3 equally likely outcomes, (g, g), (g, b), (b, g), showing that there is only a 1/3 chance that Jacobi has two girls. ■ ■

Example 4.11 Table 4.4 lists the number (in thousands) of students enrolled in a California State College, categorized by sex and age. (a) Suppose a student is randomly chosen. What is the probability this student is a woman? Find the conditional probability that a randomly chosen student is (b) (c) (d) (e)

Over 35, given that this student is a man Over 35, given that this student is a woman A woman, given that this student is over 35 A man, given that this student is between 20 and 21

Solution (a) Since there are 6663 women out of a total of 12,544 students, it follows that the probability that a randomly chosen student is a woman is 6663 = 0.5312 12,544 (b) Since there are a total of 5881 males, of whom 684 are over age 35, the desired conditional probability is P(over 35|man) =

684 = 0.1163 5881

4.5 Conditional Probability and Independence

Table 4.4 Enrollment Sex and age Total Male 14 to 17 years old 18 and 19 years old 20 and 21 years old 22 to 24 years old 25 to 29 years old 30 to 34 years old 35 years old and over Female 14 to 17 years old 18 and 19 years old 20 and 21 years old 22 to 24 years old 25 to 29 years old 30 to 34 years old 35 years old and over

12,544 5,881 91 1,309 1,089 1,080 1,016 613 684 6,663 119 1,455 1,135 968 931 716 1,339

(c) By similar reasoning to that used in (b), we see that P(over 35|woman) =

1339 = 0.2010 6663

(d) Since there are a total of 684 + 1339 = 2023 students who are over age 35, of whom 1339 are women, it follows that P(woman|over 35) =

1339 = 0.6619 2023

(e) Since there are a total of 1089 + 1135 = 2224 students who are between 20 and 21, of whom 1089 are men, it follows that P(man|between 20 and 21) =

1089 = 0.4897 2224 ■

Since P(B|A) =

P(A ∩ B) P(A)

we obtain, upon multiplying both sides by P(A), the following result, known as the multiplication rule.

171

172

CH A P T E R 4: Probability

Multiplication Rule P(A ∩ B) = P(A)P(B|A) This rule states that the probability that both A and B occur is equal to the probability that A occurs multiplied by the conditional probability of B given that A occurs. It is often quite useful for computing the probability of an intersection.

■

Example 4.12 Suppose that two people are randomly chosen from a group of 4 women and 6 men. (a) What is the probability that both are women? (b) What is the probability that one is a woman and the other a man? Solution (a) Let A and B denote, respectively, the events that the ﬁrst person selected is a woman and that the second person selected is a woman. To compute the desired probability P(A ∩ B), we start with the identity P(A ∩ B) = P(A)P(B|A) Now since the ﬁrst person chosen is equally likely to be any of the 10 people, of whom 4 are women, it follows that P(A) =

4 10

Now given that the ﬁrst person selected is a woman, it follows that the next selection is equally likely to be any of the remaining 9 people, of whom 3 are women. Therefore, P(B|A) =

3 9

and so P(A ∩ B) =

4 3 2 · = 10 9 15

(b) To determine the probability that the chosen pair consists of 1 woman and 1 man, note ﬁrst that this can occur in two disjoint ways. Either the ﬁrst person chosen is a man and the second chosen is a woman, or vice versa. Let us determine the probabilities for each of these cases. Letting A

4.5 Conditional Probability and Independence

denote the event that the ﬁrst person chosen is a man and B the event that the second person chosen is a woman, we have P(A ∩ B) = P(A)P(B|A) Now, since the ﬁrst person is equally likely to be any of the 10 people, of whom 6 are men, P(A) =

6 10

Also, given that the ﬁrst person is a man, the next selection is equally likely to be any of the remaining 9 people, of whom 4 are women, and so P(B|A) =

4 9

Therefore, P(man then woman) = P(A ∩ B) =

6 4 4 · = 10 9 15

By similar reasoning, the probability that the ﬁrst person chosen is a woman and the second chosen is a man is P(woman then man) =

4 6 4 · = 10 9 15

Since the event that the chosen pair consists of a woman and a man is the union of the above two disjoint events, we see that P(1 woman and 1 man) =

4 4 8 + = 15 15 15

■

The conditional probability that B occurs given that A has occurred is not generally equal to the (unconditional) probability of B. That is, knowing that A has occurred generally changes the chances of B’s occurrence. In the cases where P(B|A) is equal to P(B), we say that B is independent of A. Since P(A ∩ B) = P(A)P(B|A) we see that B is independent of A if P(A ∩ B) = P(A)P(B) Since this equation is symmetric in A and B, it follows that if B is independent of A, then A is also independent of B.

173

174

CH A P T E R 4: Probability

It can also be shown that if A and B are independent, then the probability of B given that A does not occur is also equal to the (unconditional) probability of B. That is, if A and B are independent, then P(B|Ac ) = P(B) Thus, when A and B are independent, any information about the occurrence or nonoccurrence of one of these events does not affect the probability of the other. Events A and B are independent if P(A ∩ B) = P(A)P(B) If A and B are independent, then the probability that a given one of them occurs is unchanged by information as to whether the other one has occurred. ■

Example 4.13 Suppose that we roll a pair of fair dice, so each of the 36 possible outcomes is equally likely. Let A denote the event that the ﬁrst die lands on 3, let B be the event that the sum of the dice is 8, and let C be the event that the sum of the dice is 7. (a) Are A and B independent? (b) Are A and C independent? Solution (a) Since A ∩ B is the event that the ﬁrst die lands on 3 and the second on 5, we see that P(A ∩ B) = P({(3, 5)}) =

1 36

On the other hand, P(A) = P({(3, 1), (3, 2), (3, 3), (3, 4), (3, 5), (3, 6)}) = and P(B) = P({(2, 6), (3, 5), (4, 4), (5, 3), (6, 2)}) = Therefore, since 1/36 = (6/36) · (5/36), we see that P(A ∩ B) = P(A)P(B) and so events A and B are not independent.

5 36

6 36

4.5 Conditional Probability and Independence

Intuitively, the reason why the events are not independent is that the chance that the sum of the dice is 8 is affected by the outcome of the ﬁrst die. In particular, the chance that the sum is 8 is enhanced when the ﬁrst die is 3, since then we still have a chance of obtaining the total of 8 (which we would not have if the ﬁrst die were 1). (b) Events A and C are independent. This is seen by noting that P(A ∩ C) = P({3, 4}) =

1 36

while P(A) =

1 6

and P(C) = P({(1, 6), (2, 5), (3, 4), (4, 3), (5, 2), (6, 1)}) =

6 36

Therefore, P(A ∩ C) = P(A)(C) and so events A and C are independent. It is rather intuitive that the event that the sum of the dice is 7 should be independent of the event that the ﬁrst die lands on 3. For no matter what the outcome of the ﬁrst die, there will always be exactly one outcome of the second die that results in the sum being equal to 7. As a result, the conditional probability that the sum is 7 given the value of the ﬁrst die will always equal 1/6. ■ ■

Example 4.14 Consider Table 4.4, presented in Example 4.11. Suppose that a female student is randomly chosen, as is, independently, a male student. Find the probability that both students are between 22 and 24 years old. Solution Since 1080 of the 5881 male students are between 22 and 24 years old, it follows that 1080 ≈ 0.1836 5881 Similarly, since 968 of the 6663 female students are between 22 and 24 years old, we see that P({male is between 22 and 24}) =

P({female is between 22 and 24}) =

968 ≈ 0.1453 6663

175

176

CH A P T E R 4: Probability

Since the choices of the male and female students are independent, we obtain P({both are between ages 22 and 24}) =

1080 968 . ≈ 0.0267 5881 6663

That is, there is approximately a 2.7 percent chance that both students are between 22 and 24 years of age. ■ While so far we have discussed independence only for pairs of events, this concept can be extended to any number of events. The probability of the intersection of any number of independent events will be equal to the product of their probabilities. If A1 , . . . , An are independent, then P(A1 ∩ A2 ∩ · · · ∩ An ) = P(A1 )P(A2 ) · · · P(An ) ■

Example 4.15 A couple is planning on having three children. Assuming that each child is equally likely to be of either sex and that the sexes of the children are independent, ﬁnd the probability that (a) All three children will be girls. (b) At least one child will be a girl. Solution (a) If we let Ai be the event that their ith child is a girl, then P(all girls) = P(A1 ∩ A2 ∩ A3 ) = P(A1 )P(A2 )P(A3 ) 1 1 1 1 = · · = 2 2 2 8

by independence

(b) The easiest way to compute the probability of at least one girl is by ﬁrst computing the probability of the complementary event—that all the children are boys. Since, by the same reasoning as used in part (a), P(all boys) =

1 8

we see that P(at least one girl) = 1 − P(all boys) =

7 8

■

4.5 Conditional Probability and Independence

PROBLEMS 1. It is estimated that 30 percent of all adults in the United States are obese, 3 percent of all adults suffer from diabetes, and 2 percent of all adults both are obese and suffer from diabetes. Determine the conditional probability that a randomly chosen individual (a) Suffers from diabetes given that he or she is obese (b) Is obese given that she or he suffers from diabetes 2. Suppose a coin is ﬂipped twice. Assume that all four possibilities are equally likely to occur. Find the conditional probability that both coins land heads given that the ﬁrst one does. 3. Consider Table 4.3 as presented in Example 4.8. Suppose that one of the workers is randomly chosen. Find the conditional probability that this worker (a) Is a woman given that he or she earns over $25,000 (b) Earns over $25,000 given that this worker is a woman 4. Fifty-two percent of the students at a certain college are females. Five percent of the students in this college are majoring in computer science. Two percent of the students are women majoring in computer science. If a student is selected at random, ﬁnd the conditional probability that (a) This student is female, given that the student is majoring in computer science (b) This student is majoring in computer science, given that the student is female Problems 5 and 6 refer to the data in the following table, which describes the age distribution of residents in a northern California county. Age 0–9 10–19 20–29 30–39 40–49 50–59 60–69 Over 70

Number 4200 5100 6200 4400 3600 2500 1800 1100

5. If a resident is randomly selected from this county, determine the probability that the resident is (a) Less than 10 years old (b) Between 10 and 20 years old

177

178

CH A P T E R 4: Probability

6.

7.

8.

9.

(c) Between 20 and 30 years old (d) Between 30 and 40 years old Find the conditional probability that a randomly chosen resident is (a) Between 10 and 20 years old, given that the resident is less than 30 years old (b) Between 30 and 40 years old, given that the resident is older than 30 A games club has 120 members, of whom 40 play chess, 56 play bridge, and 26 play both chess and bridge. If a member of the club is randomly chosen, ﬁnd the conditional probability that she or he (a) Plays chess given that he or she plays bridge (b) Plays bridge given that she or he plays chess Refer to Table 4.4, which is presented in Example 4.11. Determine the conditional probability that a randomly chosen student is (a) Less than 25 years old, given that the student is a man (b) A man, given that this student is less than 25 years old (c) Less than 25 years old, given that the student is a woman (d) A woman, given that this student is less than 25 years old Following is a pie chart detailing the after-graduation plans of the 2004 graduating class of Harvard University.

Suppose a student from this class is randomly chosen. Given that this student is not planning to go into either business or teaching, what is the probability that this student (a) Is planning to go into graduate study? (b) Is planning to go into either teaching or graduate study? (c) Is planning to go into either communications or graduate study? (d) Is not planning to go into science/technology?

4.5 Conditional Probability and Independence

(e) Is not planning to go into either communications or business? (f) Is not planning to go into either science/technology or government/politics? 10. Many psychologists believe that birth order and personality are related. To study this hypothesis, 400 elementary school children were randomly selected and then given a test to measure conﬁdence. On the results of this test each of the students was classiﬁed as being either conﬁdent or not conﬁdent. The numbers falling into each of the possible categories are: Firstborn Confident Not confident

Not ﬁrstborn

62

60

105

173

That is, for instance, out of 167 students who were ﬁrstborn children, a total of 62 were rated as being conﬁdent. Suppose that a student is randomly chosen from this group. (a) What is the probability that the student is a ﬁrstborn? (b) What is the probability that the student is rated conﬁdent? (c) What is the conditional probability that the student is rated conﬁdent given that the student is a ﬁrstborn? (d) What is the conditional probability that the student is rated conﬁdent given that the student is not a ﬁrstborn? (e) What is the conditional probability that the student is a ﬁrstborn given that the student is conﬁdent? 11. Two cards are randomly selected from a deck of 52 playing cards. What is the conditional probability they are both aces given that they are of different suits? 12. In the U.S. Presidential election of 1984, 68.3 percent of those citizens eligible to vote registered; and of those registering to vote, 59.9 percent actually voted. Suppose a citizen eligible to vote is randomly chosen. (a) What is the probability that this person voted? (b) What is the conditional probability that this person registered given that he or she did not vote? Note: In order to vote, ﬁrst you must register. 13. There are 30 psychiatrists and 24 psychologists attending a certain conference. Two of these 54 people are randomly chosen to take part in a panel discussion. What is the probability that at least one psychologist is chosen? (Hint: You may want to ﬁrst determine the probability of the complementary event that no psychologists are chosen.)

179

180

CH A P T E R 4: Probability

14. A child has 12 socks in a drawer; 5 are red, 4 are blue, and 3 are green. If 2 socks are chosen at random, ﬁnd the probability that they are (a) Both red (b) Both blue (c) Both green (d) The same color 15. Two cards are chosen at random from a deck of 52 playing cards. Find the probability that (a) Neither one is a spade (b) At least one is a spade (c) Both are spades 16. There are n socks in a drawer, of which 3 are red. Suppose that if 2 socks are randomly chosen, then the probability that they are both red is 1/2. Find n. *17. Suppose the occurrence of A makes it more likely that B will occur. In that case, show that the occurrence of B makes it more likely that A will occur. That is, show that if P(B|A) > P(B) then it is also true that P(A|B) > P(A) 18. Two fair dice are rolled. (a) What is the probability that at least one of the dice lands on 6? (b) What is the conditional probability that at least one of the dice lands on 6 given that their sum is 9? (c) What is the conditional probability that at least one of the dice lands on 6 given that their sum is 10? 19. There is a 40 percent chance that a particular company will set up a new branch ofﬁce in Chicago. If it does, there is a 60 percent chance that Norris will be named the manager. What is the probability that Norris will be named the manager of a new Chicago ofﬁce? 20. According to a geologist, the probability that a certain plot of land contains oil is 0.7. Moreover, if oil is present, then the probability of hitting it with the ﬁrst well is 0.5. What is the probability that the ﬁrst well hits oil? 21. At a certain hospital, the probability that a patient dies on the operating table during open heart surgery is 0.20. A patient who survives the operating table has a 15 percent chance of dying in the hospital from the aftereffects of the operation. What fraction of open-heart surgery patients survive both the operation and its aftereffects?

4.5 Conditional Probability and Independence

22. An urn initially contains 4 white and 6 black balls. Each time a ball is drawn, its color is noted and then it is replaced in the urn along with another ball of the same color. What is the probability that the ﬁrst 2 balls drawn are black? 23. Reconsider Prob. 7. (a) If a member is randomly chosen, what is the probability that the chosen person plays either chess or bridge? (b) How many members play neither chess nor bridge? If two members are randomly chosen, ﬁnd the probability that (c) They both play chess. (d) Neither one plays chess or bridge. (e) Both play either chess or bridge. 24. Consider Table 4.4 as given in Example 4.11. Suppose that a female student and a male student are independently and randomly chosen. (a) Find the probability that exactly one of them is over 30 years old. (b) Given that exactly one of them is over 30 years old, ﬁnd the conditional probability that the male is older. 25. José and Jim go duck hunting together. Suppose that José hits the target with probability 0.3 and Jim, independently, with probability 0.1. They both ﬁre one shot at a duck. (a) Given that exactly one shot hits the duck, what is the conditional probability that it is José’s shot? That it is Jim’s? (b) Given that the duck is hit, what is the conditional probability that José hit it? That Jim hit it? 26. A couple has two children. Let A denote the event that their older child is a girl, and let B denote the event that their younger child is a boy. Assuming that all 4 possible outcomes are equally likely, show that A and B are independent. 27. A simpliﬁed model for the movement of the price of a stock supposes that on each day the stock’s price either moves up 1 unit with probability p or moves down 1 unit with probability 1 − p. The changes on different days are assumed to be independent. Suppose that for a certain stock p is equal to 1/2. (Therefore, for instance, if the stock’s price at the end of today is 100 units, then its price at the end of tomorrow will equally likely be either 101 or 99.) (a) What is the probability that after 2 days the stock will be at its original price? (b) What is the probability that after 3 days the stock’s price will have increased by 1 unit? (c) If after 3 days the stock’s price has increased by 1 unit, what is the conditional probability that it went up on the ﬁrst day?

181

182

CH A P T E R 4: Probability

28. A male New York resident is randomly selected. Which of the following pairs of events A and B can reasonably be assumed to be independent? (a) A: He is a journalist. B: He has brown eyes. (b) A: He had a headache yesterday. B: He was in an accident yesterday.

29.

30.

31.

32.

(c) A: He is wearing a white shirt. B: He is late to work. A coin that is equally likely to land on heads or on tails is successively ﬂipped until tails appear. Assuming that the successive ﬂips are independent, what is the probability that the coin will have to be tossed at least 5 times? (Hint: Fill in the missing word in the following sentence. The coin will have to be tossed at least 5 times if the ﬁrst ______ ﬂips all land on heads.) A die is thrown until a 5 appears. Assuming that the die is equally likely to land on any of its six sides and that the successive throws are independent, what is the probability that it takes more than six throws? Suppose that the probability of getting a busy signal when you call a friend is 0.1. Would it be reasonable to suppose that the probability of getting successive busy signals when you call two friends, one right after the other, is 0.01? If not, can you think of a condition under which this would be a reasonable supposition? Two ﬁelds contain 9 and 12 plots of land, as shown here.

For an agricultural experiment, one plot from each ﬁeld will be selected at random, independently of each other. (a) What is the probability that both selected plots are corner plots? (b) What is the probability that neither plot is a corner plot? (c) What is the probability at least one of the selected plots is a corner plot? 33. A card is to be randomly selected from a deck of 52 playing cards. Let A be the event that the card selected is an ace, and let B be the event that the card is a spade. Show that A and B are independent.

4.5 Conditional Probability and Independence

34. A pair of fair dice is rolled. Let A be the event that the sum of the dice is equal to 7. Is A independent of the event that the ﬁrst die lands on 1? on 2? on 4? on 5? on 6? 35. What is the probability that two strangers have the same birthday? 36. A U.S. publication reported that 4.78 percent of all deaths in 1988 were caused by accidents. What is the probability that three randomly chosen deaths were all due to accidents? 37. Each relay in the following circuits will close with probability 0.8. If all relays function independently, what is the probability that a current ﬂows between A and B for the respective circuits? (The circuit in part (a) of the ﬁgure, which needs both of its relays to close, is called a series circuit. The circuit in part (b), which needs at least one of its relays to close, is called a parallel circuit.)

Hint: For parts (b) and (c) use the addition rule. 38. An urn contains 5 white and 5 black balls. Two balls are randomly selected from this urn. Let A be the event that the ﬁrst ball is white and B be the event that the second ball is black. Are A and B independent events? Explain your reasoning. 39. Suppose in Prob. 38 that the ﬁrst ball is returned to the urn before the second is selected. Will A and B be independent in this case? Again, explain your answer. 40. Suppose that each person who is asked whether she or he is in favor of a certain proposition will answer yes with probability 0.7 and no with probability 0.3. Assume that the answers given by different people are independent. Of the next four people asked, ﬁnd the probability that (a) All give the same answer. (b) The ﬁrst two answer no and the ﬁnal two yes. (c) At least one answers no. (d) Exactly three answer yes. (e) At least one answers yes. 41. The following data, obtained from the U.S. National Oceanic and Atmospheric Administration, give the average number of days with precipitation of 0.01 inch or more in different months for the cities of Mobile, Phoenix, and Los Angeles.

183

184

CH A P T E R 4: Probability

Average Number of Days with Precipitation of 0.01 Inch or More City Mobile Phoenix Los Angeles

January

April

July

11 4 6

7 2 3

16 4 1

Suppose that in the coming year you are planning to visit Phoenix on January 4, Los Angeles on April 10, and Mobile on July 15. (a) What is the probability that it will rain on all three trips? (b) What is the probability it will be dry on all three trips? (c) What is the probability that you encounter rain in Phoenix and Mobile but not in Los Angeles? (d) What is the probability that you encounter rain in Mobile and Los Angeles but not in Phoenix? (e) What is the probability that you encounter rain in Phoenix and Los Angeles but not in Mobile? (f) What is the probability that it rains in exactly two of your three trips? 42. Each computer chip produced by machine A is defective with probability 0.10, whereas each chip produced by machine B is defective with probability 0.05. If one chip is taken from machine A and one from machine B, ﬁnd the probability (assuming independence) that (a) Both chips are defective. (b) Both are not defective. (c) Exactly one of them is defective. If it happens that exactly one of the two chips is defective, ﬁnd the probability that it was the one from (d) Machine A (e) Machine B 43. Genetic testing has enabled parents to determine if their children are at risk for cystic ﬁbrosis (CF), a degenerative neural disease. A child who receives a CF gene from both parents will develop the disease by his or her teenage years and will not live to adulthood. A child who receives either zero or one CF gene will not develop the disease; however, if she or he does receive one CF gene, it may be passed on to subsequent offspring. If an individual has a CF gene, then each of his or her children will receive that gene with probability 1/2. (a) If both parents possess the CF gene, what is the probability that their child will develop cystic ﬁbrosis? (b) What is the probability that a 25-year-old person who does not have CF but whose sibling does, carries the gene?

4.6 Bayes’ Theorem

*4.6 BAYES’ THEOREM For any two events A and B, we have the following representation for A: A = (A ∩ B) ∪ (A ∩ Bc ) That this is valid is easily seen by noting that for an outcome to be in A, either it must be in both A and B or it must be in A but not in B (see Fig. 4.7). Since A ∩ B and A ∩ Bc are mutually exclusive (why?), we have by Property 3 (see Sec. 4.3) P(A) = P(A ∩ B) + P(A ∩ Bc ) Since P(A ∩ B) = P(A|B)P(B)

and

P(A ∩ Bc ) = P(A|Bc )P(Bc )

we have thus shown the following equality: P(A) = P(A|B)P(B) + P(A|Bc )P(Bc )

(4.1)

This equality states that the probability of event A is a weighted average of the conditional probability of A given that B occurs and the conditional probability of A given that B does not occur; each conditional probability is weighted by the probability of the event on which it is conditioned. It is a very useful formula for it often enables us to compute the probability of an event A by ﬁrst “conditioning” on whether a second event B occurs. Before illustrating the use of Eq. (4.1), we ﬁrst consider the problem of how to reevaluate an initial probability in light of additional evidence. Suppose there is a certain hypothesis under consideration; let H denote the event that the hypothesis is true and P(H) the probability that the hypothesis is true. Now, suppose that additional evidence, call it E, concerning this hypothesis becomes available. We thus want to determine P(H|E), the conditional probability that the

FIGURE 4.7 A = (A ∩ B) ∪ (A ∪ Bc )

185

186

CH A P T E R 4: Probability

hypothesis is true given the new evidence E. Now, by the deﬁnition of conditional probability, P(H|E) =

P(H ∩ E) P(E|H)P(H) = P(E) P(E)

By making use of Eq. (4.1), we can compute P(E) by conditioning on whether the hypothesis is true. This yields the following identity, known as Bayes’ theorem. Bayes’ Theorem P(H|E) =

■

P(E|H)P(H) P(E|H)P(H) + P(E|H c )P(H c )

Example 4.16 An insurance company believes that people can be divided into two classes— those who are prone to have accidents and those who are not. The data indicate that an accident-prone person will have an accident in a 1-year period with probability 0.1; the probability for all others is 0.05. Suppose that the probability is 0.2 that a new policyholder is accident-prone. (a) What is the probability that a new policyholder will have an accident in the ﬁrst year? (b) If a new policyholder has an accident in the ﬁrst year, what is the probability that he or she is accident-prone? Solution Let H be the event that the new policyholder is accident-prone, and let A denote the event that she or he has an accident in the ﬁrst year. We can compute P(A) by conditioning on whether the person is accident-prone: P(A) = P(A|H)P(H) + P(A|H c )P(H c ) = (0.1)(0.2) + (0.05)(0.8) = 0.06 Therefore, there is a 6 percent chance that a new policyholder will have an accident in the ﬁrst year. We compute P(H|A) as follows: P(H ∩ A) P(A) P(A|H)P(H) = P(A) 1 (0.1)(0.2) = = 0.06 3

P(H|A) =

4.6 Bayes’ Theorem

Therefore, given that a new policyholder has an accident in the ﬁrst year, the conditional probability that the policyholder is prone to accidents is 1/3. ■ ■

Example 4.17 A blood test is 99 percent effective in detecting a certain disease when the disease is present. However, the test also yields a false-positive result for 2 percent of the healthy patients tested. (That is, if a healthy person is tested, then with probability 0.02 the test will say that this person has the disease.) Suppose 0.5 percent of the population has the disease. Find the conditional probability that a randomly tested individual actually has the disease given that his or her test result is positive. Solution Let D denote the event that the person has the disease, and let E be the event that the test is positive. We want to determine P(D|E), which can be accomplished by using Bayes’ theorem as follows: P(D|E) = =

P(E|D)P(D) P(E|D)P(D) + P(E|Dc )P(Dc ) (0.99)(0.005) = 0.199 (0.99)(0.005) + (0.02)(0.995)

Thus, there is approximately a 20 percent chance that a randomly chosen person from the population who tests positive actually has the disease. (The reason why it is so low is that the chance that a randomly chosen person is free of the disease yet tests positive is greater than the chance that the person has the disease and tests positive.) ■

PROBLEMS 1. There are two coins on a table. When both are ﬂipped, one coin lands on heads with probability 0.5 while the other lands on heads with probability 0.6. A coin is randomly selected from the table and ﬂipped. (a) What is the probability it lands on heads? (b) Given that it lands on tails, what is the conditional probability that it was the fair coin (that is, the one equally likely to land heads or tails)? 2. Suppose that when answering a question on a multiple-choice test, a student either knows the answer or guesses at it. If he guesses at the

187

188

CH A P T E R 4: Probability

3.

4.

5.

6.

7.

8.

answer, then he will be correct with probability 1/5. If the probability that a student knows the answer is 0.6, what is the conditional probability that the student knew the answer given that he answered it correctly? The inspector in charge of a criminal investigation is 60 percent certain of the guilt of a certain suspect. A new piece of evidence proving that the criminal was left-handed has just been discovered. Whereas the inspector knows that 18 percent of the population is left-handed, she is waiting to ﬁnd out whether the suspect is left-handed. (a) What is the probability that the suspect is left-handed? (b) If the suspect turns out to be left-handed, what is the probability that the suspect is guilty? Urn 1 contains 4 red and 3 blue balls, and urn 2 contains 2 red and 2 blue balls. A ball is randomly selected from urn 1 and placed in urn 2. A ball is then drawn from urn 2. (a) What is the probability that the ball drawn from urn 2 is red? (b) What is the conditional probability that the ball drawn from urn 1 is red given that a blue ball is drawn from urn 2? Consider a diagnostic test that is 97 percent accurate on both those who have and those who do not have the disease. (That is, if a person has the disease, then with probability 0.97 the diagnosis will be positive; and if the person does not have the disease, then with probability 0.97 the diagnosis will be negative.) Suppose 2 percent of the population has the disease. What is the conditional probability that a randomly selected member of the population has the disease if that person’s diagnosis was positive? There are three cards in a hat. One is colored red on both sides, one is black on both sides, and one is red on one side and black on the other. The cards are thoroughly mixed in the hat, and one card is drawn and placed on a table. If the side facing up is red, what is the conditional probability that the other side is black? A total of 52 percent of voting-age residents of a certain city are Republicans, and the other 48 percent are Democrats. Of these residents, 64 percent of the Republicans and 42 percent of the Democrats are in favor of discontinuing afﬁrmative action city hiring policies. A voting-age resident is randomly chosen. (a) What is the probability that the chosen person is in favor of discontinuing afﬁrmative action city hiring policies? (b) If the person chosen is against discontinuing afﬁrmative action hiring policies, what is the probability she or he is a Republican? A person’s eye color is determined by a single pair of genes. If both genes are blue-eyed genes, then the person will have blue eyes; if they are both brown-eyed genes, then the person will have brown eyes;

4.7 Counting Principles

and if one gene is blue-eyed and the other is brown-eyed, then the person will have brown eyes. (Because of this latter fact we say that the brown-eyed gene is dominant over the blue-eyed one.) A newborn child independently receives one eye gene from each parent, and the gene that the child receives from a parent is equally likely to be either of the two eye genes of that parent. Suppose that Susan has blue eyes and both her parents have brown eyes. (a) What is the eye gene pair of Susan’s mother? of her father? (b) Susan’s brown-eyed sister is pregnant. If her sister’s husband has blue eyes, what is the probability the baby will have blue eyes? Hint: What is the probability that Susan’s sister has a blue-eyed gene? 9. Twelve percent of all U.S. households are in California. A total of 1.3 percent of all U.S. households earn over 250,000 dollars per year, while a total of 3.3 percent of all California households earn over 250,000 dollars per year. A U.S. household is randomly chosen. (a) What percentage of non-California households earn over 250,000 dollars per year? (b) Given that the chosen household earns over 250,000 dollars per year, what is the probability it is a California household?

*4.7 COUNTING PRINCIPLES As seen in Sec. 4.4, we often determine probabilities by counting the number of different outcomes in a speciﬁed event. The key to doing this effectively is to make use of the following rule, known as the basic principle of counting. Basic Principle of Counting Suppose an experiment consists of two parts. If part 1 can result in any of n possible outcomes and if for each outcome of part 1 there are m possible outcomes of part 2, then there is a total of nm possible outcomes of the experiment. That the basic principle is valid can easily be seen by enumerating all possible outcomes of the experiment: (1, 1),

(1, 2),

...,

(1, m)

(2, 1), . . .

(2, 2),

...,

(2, m)

(n, 1),

(n, 2),

...,

(n, m)

189

190

CH A P T E R 4: Probability

where we say that the outcome of the experiment is (i, j) if part 1 of the experiment resulted in its ith possible outcome and part 2 then resulted in its jth possible outcome. Since the preceding display contains n rows, each of which consists of m outcomes, it follows that there is total of m + m · · · + m = nm outcomes. ■

Example 4.18 One man and one woman are to be selected from a group consisting of 12 women and 8 men. How many different choices are possible? Solution By regarding the choice of the woman as the ﬁrst part of the experiment and the choice of the man as the second, we see from the basic principle that there are 12 · 8 = 96 possible outcomes. ■

■

Example 4.19 Two people are to be selected from a group that consists of 10 married couples. How many different choices are possible? If each choice is equally likely, what is the probability that the two people selected are married to each other? Solution Since the ﬁrst person selected is any of the 20 people and the next one is then any of the remaining 19, it follows from the basic principle that there are 20 · 19 = 380 possible outcomes. Now, for each married couple there are 2 outcomes that result in that couple’s selection. Namely, the husband could be the ﬁrst person selected and the wife the second, or the reverse. Thus, there are 2 · 10 = 20 different outcomes that result in a married couple’s selection. Hence, assuming that that all possible outcomes are equally likely, it follows that the probability that the people selected are married to each other ■ is 20/380 = 1/19.

When the experiment consists of more than two parts, the basic principle can be generalized as follows. Generalized Basic Principle of Counting Suppose an experiment consists of r parts. Suppose there are n1 possible outcomes of part 1 and then n2 possible outcomes of part 2 and then n3 possible outcomes of part 3, and so on. Then there is a total of n1 · n2 · · · nr possible outcomes of the experiment. As an application of the generalized principle, suppose that we want to determine the number of different ways that the three letters a, b, c can be arranged in a linear

4.7 Counting Principles

order. By direct enumeration we can see that there are 6 possible arrangements: abc, acb, bac, bca, cab, cba This result could also have been obtained by using the generalized basic principle of counting. That is, there are 3 choices for the ﬁrst element in the ordering, there are then 2 choices for the second, and then 1 choice for the third position. Hence, there are 3 · 2 · 1 = 6 possible outcomes. Suppose now that we want to determine the number of different arrangements of n objects. By the same reasoning, we see that there is a total of n · (n − 1) · (n − 2) · · · 3 · 2 · 1 different arrangements. Each of these arrangements is called a permutation. It is convenient to introduce for the foregoing expression the notation n!, which is read “n factorial.” That is, n! = n · (n − 1) · (n − 2) · · · 3 · 2 · 1 Thus, for instance, 1! = 1 2! = 2 · 1 = 2 3! = 3 · 2 · 1 = 6 4! = 4 · 3 · 2 · 1 = 24 and so on. In addition, it is convenient to deﬁne 0! to be equal to 1. ■

Example 4.20 If four people are in a room, what is the probability that no two of them celebrate their birthday on the same day of the year? Solution Since each person can celebrate his or her birthday on any of the 365 days of the year, it follows from the generalized basic principle that there is a total of 365 · 365 · 365 · 365 = (365)4 possible outcomes. (We are ignoring the possibility that someone was born on February 29.) Let us now determine the number of outcomes in which no two individuals have the same birthday. This will occur if the birthday of the ﬁrst person is any of the 365 days, the birthday of the second person is then any one of the remaining 364 days, the birthday of the third person is then any one

191

192

CH A P T E R 4: Probability

of the remaining 363 days, and the birthday of the ﬁnal person is then any of the remaining 362 days. Thus, by the generalized basic principle of counting, we see that there is a total of 365 · 364 · 363 · 362 different outcomes in which all the 4 birthdays are different. Hence, if we assume that all of the possible outcomes are equally likely, we obtain the probability that no two people have the same birthday: 365 · 364 · 363 · 362 = 0.983644 365 · 365 · 365 · 365 The same approach can be used to ﬁnd the probability that a group of n people will all have different birthdays, for any integer n. It is an interesting fact that when n = 23, this probability is less than 1/2. That is, if there are 23 people in a room, then it is more likely than not that at least two of them will celebrate the same birthday. ■ Suppose now that we are interested in choosing 3 of the 5 items a, b, c, d, e. How many different choices are possible? To answer this we can reason as follows. Since there are 5 possible choices for the ﬁrst item and then 4 possible choices for the next one and ﬁnally 3 possible choices for the ﬁnal item, it follows that there are 5 · 4 · 3 possible choices when the order in which the items are chosen is considered relevant. However, in this set of ordered choices, every group of three items will appear 3! times. For instance, consider the group consisting of the items a, b, and c. Each of the permutations abc, acb, bac, bca, cab, cba of these three elements will be included in the set of possible choices when the order of selection is considered relevant. Therefore it follows that the number of different groups of size 3 that can be formed from 5 items, when the order of selection is not considered relevant, is 5·4·3 = 10 3·2·1 Suppose now that we are interested in determining the number of different groups of size r that can be chosen from a set of n elements. By the same reasoning as before, it follows that there are n · (n − 1) · · · (n − r + 1) r! different groups. Since, n(n − 1) · · · (n − r + 1) can be written as n!/(n − r)!, we can express this number as n!/[(n − r)! r! ].

4.7 Counting Principles

Notation and Terminology Deﬁne nr , for r ≤ n, by n n! n(n − 1) · · · (n − r + 1) = = r (n − r)! r! r! Call nr the number of combinations of n things taken r at a time; it represents the number of different groups of size r that can be selected from a set of size n when the order of selection is not of importance. ■

Example 4.21 (a) How many different groups of size 2 can be selected from the items a, b, c? (b) How many different groups of size 2 can be chosen from a set of 6 people? (c) How many different groups of size 3 can be chosen from a set of 6 people? Solution

3·2 (a) There are 32 = 2·1 = 3 different groups of 2 items that can be selected from the items a, b, c: a and b, a and c, and b and c.

(b) and (c) From a set of 6 people there are 6 6·5 = 15 = 2·1 2 different groups of size 2 that can be chosen, and 6 6·5·4 = 20 = 3·2·1 3 different groups of size 3. ■

■

Example 4.22 A random sample of size 3 is to be selected from a set of 10 items. What is the probability that a prespeciﬁed item will be selected? Solution

There are 10 3 different groups that can be chosen. The number of different groups that contain the speciﬁed item is equal to the number of choices of the additional 2 items from the remaining 9 items after the speciﬁed item is chosen. Thus, there are 92 different groups that contain the given item. So, assuming that a random sample is one in which each group is equally likely to be selected,

193

194

CH A P T E R 4: Probability

we see that the desired probability that a given item is selected is 9 9·8·3·2·1 3 9 · 8 10 · 9 · 8 2 10 ÷ = = = 2·1 3·2·1 2 · 1 · 10 · 9 · 8 10 3 That is, there is a 3-in-10 chance that a given item will be selected. ■

■

Example 4.23 A committee of 4 people is to be selected from a group of 5 men and 7 women. If the selection is made randomly, what is the probability the committee will consist of 2 men and 2 women? Solution will assume that “the selection is made randomly” means that each of the 5 We 12 4 possible combinations is equally 7 likely to be chosen. Because there are 2 possible choices of 2 men and 2 possible choices of 2 women, it follows from the basic principle of counting that there are 52 72 possible outcomes that contain 2 men and 2 women. Therefore, the desired probability is 57 14 5·4·7·6·4·3·2·1 2 122 = = ■ 2 · 1 · 2 · 1 · 12 · 11 · 10 · 9 33 4

It follows from the formula

that

■

n! n = r!(n − r)! r n n = r n−r

Example 4.24 Compare Solution

8 5

and

12 10

. 8 8 8·7·6 = = = 56 5 3 3·2·1 12 12 12 · 11 = = = 66 10 2 2·1

■

n The identity nr = n−r can be seen by a “counting argument.” Suppose we want to select r items from a set of n items. Since this can be done either by directly

4.7 Counting Principles

specifying the r items to be selected or equivalently by specifying the n − r items that are not to be selected, it follows that the number of choices of r items is equal to the number of choices of n − r items. For instance, any choice of 8 of the ﬁrst 10 integers corresponds to a choice of the 2 integers that are not chosen. ■

Example 4.25 Suppose that n + m digits, n of which are equal to 1 and m of which are equal to 0, are to be arranged in a linear order. How many different arrangements are possible? For instance, if n = 2 and m = 1, then there are 3 possible arrangements: 1, 1, 0 1, 0, 1 0, 1, 1 Solution Each arrangement will have a digit in position 1, another digit in position 2, another in position 3, . . . , and ﬁnally a digit in position n + m. Each arrangement can therefore be described by specifying the n positions that contain the digit 1. That is, each different choice of n ofthe n + m positions will result in a different arrangements. different arrangement. Therefore, there are n+m n Of course, we can also describe an arrangement by specifying the m positions that contain the digit 0. This results in the solution n+m m , which is equal to n+m ■ n .

PROBLEMS 1. How many different 7-place license plates are possible when the ﬁrst 3 places are for letters and the last 4 are for digits? 2. How many different batting orders are possible for a baseball team consisting of 9 players? 3. 9! = 362,880. What is the value of 10!? 4. There is a certain type of combination lock that has a dial that can be stopped at any of the numbers 1 through 36. To open the lock you have to twirl the dial clockwise until a certain number is reached, then twirl it counterclockwise until a second number is reached, and then twirl it clockwise until a third number is reached. If you have forgotten the three numbers (which need not be different from each other), how many different possibilities might you have to try before the lock opens? 5. Telephone area codes in the United States and Canada consist of a sequence of three digits. The ﬁrst digit is an integer between 2 and 9; the second digit is either 0 or 1; the third digit is any integer between

195

196

CH A P T E R 4: Probability

1 and 9. How many area codes are possible? How many area codes starting with a 4 are possible? 6. A well-known nursery tale starts as follows: As I was going to St. Ives I met a man with 7 wives. Each wife had 7 sacks, Each sack had 7 cats, Each cat had 7 kittens. How many kittens did our traveler meet? 7. (a) If four workers are to be assigned to four jobs how many different assignments are possible? (b) How many assignments are possible if workers 1 and 2 are both qualiﬁed only for jobs 1 and 2 and workers 3 and 4 are both qualiﬁed only for jobs 3 and 4? 8. Use the formula n n! = r (n − r)! r! to ﬁnd 0n , where n n is a positive integer. Recall that 0! is deﬁned to equal 1. Since r is supposed to equal the number of groups of size r that can be formed from a set of n objects, do you think the answer makes sense? 9. Calculate the following: 8 9 7 10 , , , 4 2 6 3 10. Consider a group of 20 people. If everyone shakes hands with everyone else, how many handshakes take place? 11. A student must choose four courses from among French, Spanish, History, Physics, and English Literature. (a) How many different choices are possible? (b) If the student chooses randomly, what is the probability that both French and Spanish are chosen? 12. A delivery company has 10 trucks, of which 3 have faulty brakes. If an inspector randomly chooses 2 of the trucks for a brake check, what is the probability that none of the trucks with faulty brakes are chosen? 13. A company regularly receives large shipments of computer chips. The company’s policy is to randomly select and test 10 of the chips. If 2 or more of these are found to be defective, then the shipment is returned; otherwise the shipment is accepted. Suppose that a shipment of 100 chips contains 14 that are defective.

4.7 Counting Principles

14.

15.

16.

17.

18.

(a) What is the probability that the sample inspected has no defective chips? (b) What is the probability that the sample inspected has 1 defective chip? (c) What is the probability this shipment will be rejected? In a state lottery, a player must choose 8 of the numbers from 1 to 40. The Lottery Commission then performs an experiment that selects 8 of these 40 numbers. Assuming that the choice of the Lottery Commis sion is equally likely to be any of the 408 combinations, what is the probability that a player has (a) All 8 of the selected numbers? (b) Seven of the selected numbers? (c) At least 6 of the selected numbers? An approved jury list contains 22 men and 18 women. What is the probability that a random selection of 12 of these people will result in a jury with (a) Six women and 6 men? (b) Eight women and 4 men? (c) At least 10 men? The second Earl of Yarborough is reported to have bet at odds of 1000 to 1 that a bridge hand of 13 cards would contain at least one card that is 10 or higher. (By 10 or higher we mean that it is either ten, jack, queen, king, or ace.) Nowadays, we call a hand that has no cards higher than 9 a Yarborough. What is the probability that a randomly selected bridge hand is a Yarborough? An instructor gives her class a set of 10 problems and tells the class that the ﬁnal exam (in 1 week) will consist of a random selection of 5 of the problems. If a student has ﬁgured out how to do 7 of the problems by the time of the exam, what is the probability he or she will correctly answer (a) All 5 problems? (b) At least 4 of the problems? Consider the grid of points shown here.

Suppose that starting at the point labeled A you can at each move either go one step up or one step to the right. You keep doing this until the point labeled B is reached. How many different paths from

197

198

CH A P T E R 4: Probability

A to B are possible? Hint: To go from A to B you have to go 4 steps to the right and 3 steps up. Indeed, any path can be speciﬁed by an arrangement of 4 r’s and 3 u’s. For instance, the arrangement r, r, r, r, u, u, u speciﬁes the following path:

19. Suppose, in Prob. 18, that a path from A to B is randomly chosen. What is the probability it goes through the point circled in the following grid? (Hint: How many paths are there from A to the circled point? How many from the circled point to B?)

KEY TERMS Experiment: Any process that produces an observation. Outcome: The observation produced by an experiment. Sample space: The set of all possible outcomes of an experiment. Event: Any set of outcomes of the experiment. An event is a subset of sample space S. The event is said to occur if the outcome of the experiment is contained in it. Union of events: The union of events A and B, denoted by A ∪ B, consists of all outcomes that are in A or in B or in both A and B. Intersection of events: The intersection of events A and B, denoted by A ∩ B, consists of all outcomes that are in both A and B. Complement of an event: The complement of event A, denoted by Ac , consists of all outcomes that are not in A. Mutually exclusive or disjoint: Events are mutually exclusive or disjoint if they cannot occur simultaneously.

Key Terms

Null event: The event containing no outcomes. It is the complement of sample space S. Venn diagram: A graphical representation of events. Probability of an event: The probability of event A, denoted by P(A), is the probability that the outcome of the experiment is contained in A. Addition rule of probability: The formula P(A ∪ B) = P(A) + P(B) − P(A ∩ B) Conditional probability: The probability of one event given the information that a second event has occurred. We denote the conditional probability of B given that A has occurred by P(B|A). Multiplication rule: The formula P(A ∩ B) = P(A)P(B|A) Independent: Two events are said to be independent if knowing whether a speciﬁc one has occurred does not change the probability that the other occurs.

199

200

CH A P T E R 4: Probability

SUMMARY Let S denote all possible outcomes of an experiment whose outcome is not predictable in advance. S is called the sample space of the experiment. Any set of outcomes, or equivalently any subset of S, is called an event. If A and B are events, then A ∪ B, called the union of A and B, is the event consisting of all outcomes that are in A or in B or in both A and B. The event A ∩ B is called the intersection of A and B. It consists of all outcomes that are in both A and B. For any event A, we deﬁne the event Ac , called the complement of A, to consist of all outcomes in S that are not in A. The event Sc , which contains no outcomes, is designated by ∅. If A ∩ B = ∅, meaning that A and B have no outcomes in common, then we say that A and B are disjoint (also called mutually exclusive). We suppose that for every event A there is a number P(A), called the probability of A. These probabilities satisfy the following three properties. PROPERTY 1: 0 ≤ P(A) ≤ 1 PROPERTY 2: P(S) = 1 PROPERTY 3: P(A ∪ B) = P(A) + P(B) when A ∩ B = ∅ The quantity P(A) represents the probability that the outcome of the experiment is in A. If so, we say that A occurs. The identity P(A ∪ B) = P(A) + P(B) − P(A ∩ B) is called the addition rule of probability. We sometimes assume that all the outcomes of an experiment are equally likely. Under this assumption, it can be shown that P(A) =

Numbers of outcomes in A Numbers of outcomes in S

The conditional probability of B given that A has occurred is denoted by P(B|A) and is given by the following equation: P(B|A) =

P(A ∩ B) P(A)

Multiplying both sides of this equation by P(A) gives the following identity, known as the multiplication rule: P(A ∩ B) = P(A)P(B|A)

Review Problems

If P(A ∩ B) = P(A)P(B) then we say that events A and B are independent. If A and B are independent, then the probability that one of them will occur is unchanged by information as to whether the other has occurred.

REVIEW PROBLEMS 1. Of 12 bottles in a case of wine, 3 are bad. Suppose 2 bottles are randomly chosen from the case. Find the probability that (a) The ﬁrst bottle chosen is good. (b) The second bottle chosen is good. (c) Both bottles are good. (d) Both bottles are bad. (e) One is good, and one is bad. 2. A basketball player makes each of her foul shots with probability 0.8. Suppose she is fouled and is awarded two foul shots. Assuming that the results of different foul shots are independent, ﬁnd the probability that she (a) Makes both shots (b) Misses both shots (c) Makes the second shot given that she missed the ﬁrst 3. Suppose that a basketball player makes her ﬁrst foul shot with probability 0.8. However, suppose that the probability that she makes her second shot depends on whether the ﬁrst shot is successful. Speciﬁcally, suppose that if she is successful on her ﬁrst shot, then her second will be successful with probability 0.85, whereas if she misses her ﬁrst shot, then the second will be successful with probability 0.7. Find the probability that she (a) Makes both shots (b) Misses both shots (c) Makes the ﬁrst but misses the second shot 4. Of the registered voters in a certain community 54 percent are women and 46 percent are men. Sixty-eight percent of the registered women voters and 62 percent of the registered men voters voted in the last local election. If a registered voter from this community is randomly chosen, ﬁnd the probability that this person is (a) A woman who voted in the last election (b) A man who did not vote in the last election (c) What is the conditional probability that this person is a man given that this person voted in the last election?

201

202

CH A P T E R 4: Probability

5. A kindergarten class consists of 24 students—13 girls and 11 boys. Each day one of these children is chosen as “student of the day.” The selection is made as follows. At the beginning of the school year, the names of the children are written on slips of paper, which are then put in a large urn. On the ﬁrst day of school, the urn is shaken and a name is chosen to be student of the day. The next day this process is repeated with the remaining 23 slips of paper, and so on. When each student has been selected once (which occurs on day 24), the process is repeated. (a) What is the probability that the ﬁrst selection is a boy? (b) If the ﬁrst selection is a boy, what is the probability that the second is a girl? 6. Two cards are chosen from an ordinary deck of 52 playing cards. Find the probability that (a) Both are aces. (b) Both are spades. (c) They are of different suits. (d) They are of different denominations. 7. What is the probability of the following outcomes when a fair coin is independently tossed 6 times? (a) H H H H H H (b) H T H T H T (c) T T H H T H 8. Find the probability of getting a perfect score just by guessing on a true/false test with (a) 2 questions (b) 3 questions (c) 10 questions 9. A cafeteria offers a three-course meal. One chooses a main course, a starch, and a dessert. The possible choices are given in this table. Meal Main course Starch course Dessert

Choices Chicken or roast beef Rice or potatoes Melon or ice cream or gelatin

Let the outcome of an experiment be the dinner selection of a person who makes one selection from each of the courses. (a) List all the outcomes in sample space S. (b) Suppose the person is allergic to rice and melon. List all the outcomes in the event corresponding to a choice that is acceptable to this person.

Review Problems

(c) If the person randomly chooses a dessert, what is the probability it is ice cream? (d) If the person makes a random choice in each of the courses, what is the probability that chicken, rice, and melon are chosen? 10. The following is a breakdown by age and sex of the population of the United States. The numbers in each class are in units of 1 million. Sex Age

Females

Males

48.8 74.5

50.4 66.6

Under 25 years Over 25 years

11.

12.

13.

14.

Suppose a person is chosen at random. Let A be the event that the person is male and B be the event that the person is under age 25. Find (a) P(A) and P(Ac ) (b) P(B) and P(Bc ) (c) P(A ∩ B) (d) P(A ∩ Bc ) (e) P(A|B) (f) P(B|A) A person has three keys of which only one ﬁts a certain lock. If she tries the keys in a random order, ﬁnd the probability that (a) The successful key is the ﬁrst one tried. (b) The successful key is the second one tried. (c) The successful key is the third one tried. (d) The second key works given that the ﬁrst one did not. Two cards from an ordinary playing deck constitute a blackjack if one card is an ace and the other is a face card, where a face card is 10, jack, queen, or king. What is the probability that a random selection of two cards yields a blackjack? (Hint: You might try to compute the probability that the ﬁrst card is an ace and the second a face card, and the probability that the ﬁrst is a face card and the second an ace.) A delivery company has 12 trucks, of which 4 have faulty brakes. If an inspector randomly chooses 2 of the trucks for a brake check, what is the probability that neither one has faulty brakes? Suppose that A and B are independent events, and P(A) = 0.8 Find (a) P(A ∩ B) (b) P(A ∪ B) (c) P(B) (d) P(Ac ∩ B)

P(Bc ) = 0.4

203

204

CH A P T E R 4: Probability

15. A deck of 52 cards is shufﬂed, and the cards are turned face up, one at a time. (a) What is the probability that the ﬁrst card turned up is the ace of spades? (b) Let A denote the event that the ﬁrst card turned up is not the ace of spades, and let B denote the event that the second card turned up is the ace of spades. Therefore, A ∩ B is the event that the second card turned up is the ace of spades. Compute the probability of this event by using P(A ∩ B) = P(A)P(B|A) (c) Fill in the missing word in the following intuitive argument for the solution obtained in part (b): Since all orderings are equally likely, the second card turned up is _______ likely to be any of the 52 cards. (d) What is the probability that the 17th card turned up is the ace of spades? 16. Floppy disks go through a two-stage inspection procedure. Each disk is checked ﬁrst manually and then electronically. If the disk is defective, then a manual inspection will spot the defect with probability 0.70. A defective disk that passes the manual inspection will be detected electronically with probability 0.80. What percentage of defective disks is not detected? 17. Assume that business conditions in any year can be classiﬁed as either good or bad. Suppose that if business is good this year, then with probability 0.7 it will also be good next year. Also suppose that if business is bad this year, then with probability 0.4 it will be good next year. The probability that business will be good this year is 0.6. Find the probabilities that the following statements are true. (a) Business conditions both this year and next will be good. (b) Business conditions will be good this year and bad next year. (c) Business conditions will be bad both years. (d) Business conditions will be good next year. (e) Given that business conditions are good next year, what is the conditional probability that they were good this year? 18. Both John and Maureen have one gene for blue eyes and one for brown eyes. A child of theirs will receive one gene for eye color from Maureen and one from John. The gene received from each parent is equally likely to be either of the parent’s two genes. Also, the gene received from John is independent of the one received from Maureen. If a child receives a blue gene from both John and Maureen, then that child will have blue eyes; otherwise, it will have brown eyes. Maureen and John have two children.

Review Problems

(a) What is the probability that their older child has blue eyes? (b) What is the probability that the older child has blue and the younger has brown eyes? (c) What is the probability that the older has brown and the younger has blue eyes? (d) What is the probability that one child has blue eyes and the other has brown eyes? (e) What is the probability they both have blue eyes? (f) What is the probability they both have brown eyes? 19. It is estimated that for the U.S. adult population as a whole, 55 percent are above ideal weight, 20 percent have high blood pressure, and 60 percent either are above ideal weight or have high blood pressure. Let A be the event that a randomly chosen member of the population is above his or her ideal weight, and let B be the event that this person has high blood pressure. Are A and B independent events? 20. A card is randomly selected from a deck of playing cards. Let A be the event that the card is an ace, and let B be the event that it is a spade. State whether A and B are independent, if the deck is (a) A standard deck of 52 cards (b) A standard deck, with all 13 hearts removed (c) A standard deck, with the hearts from 2 through 9 removed 21. A total of 500 married working couples were polled about whether their annual salaries exceeded $75,000. The following information was obtained:

Husband Wife

Less than $75,000

More than $75,000

Less than $75,000

212

198

More than $75,000

36

54

Thus, for instance, in 36 couples, the wife earned over $75,000 and the husband earned less than $75,000. One of the couples is randomly chosen. (a) What is the probability that the husband earns less than $75,000? (b) What is the conditional probability that the wife earns more than $75,000 given that the husband earns more than this amount? (c) What is the conditional probability that the wife earns more than $75,000 given that the husband earns less than this amount? (d) Are the salaries of the wife and husband independent? 22. The probability that a new car battery functions for over 10,000 miles is 0.8, the probability it functions for over 20,000 miles is 0.4, and the

205

206

CH A P T E R 4: Probability

probability it functions for over 30,000 miles is 0.1. If a new car battery is still working after 10,000 miles, ﬁnd the conditional probability that (a) Its total life will exceed 20,000 miles. (b) Its additional life will exceed 20,000 miles. 23. Of the drivers who stop at a certain gas station, 90 percent purchase either gasoline or oil. A total of 86 percent purchase gasoline, and 8 percent purchase oil. (a) What percentage of drivers purchase gasoline and oil? (b) Find the conditional probability that a driver (i) Purchases oil given that she or he purchases gasoline (ii) Purchases gasoline given that he or she purchases oil (iii) Suppose a driver stops at a gas station. Are the events that the driver purchases oil and that the driver purchases gasoline independent? The following table gives participation rates at various artistic and leisure activities for individuals in different age categories. The data are for the year 2000, and the numbers represent the proportion of the population being considered who satisﬁed the stated criteria. Attended at least once

Characteristic Average 18–24 years old 25–34 years old 35–44 years old 45–54 years old 55–64 years old 65–74 years old 75 years old and over Male Female

Jazz performance

Classical music performance

Opera performance

Musical plays

Plays

Ballet performance

10 14 15 10 8 5 3 1 10 9

13 11 12 16 15 11 13 10 11 14

3 2 2 4 4 3 3 1 2 3

17 15 16 21 20 18 13 8 15 19

12 11 12 14 13 10 10 7 11 12

4 4 5 6 3 4 4 2 3 5

Visited at least once— art museum or gallery

Read— novel, short stories, poetry, or plays

22 22 26 27 22 19 16 10 21 23

56 57 59 62 57 50 50 48 48 63

Source: U.S. National Endowment for the Arts.

Problems 24 to 26 refer to the preceding table. 24. Suppose an 18- to 24-year-old is randomly chosen, as is a 35- to 44year-old. Find the probability that (a) Both attended a jazz performance. (b) Exactly one attended a jazz performance.

Review Problems

25.

26.

27.

28.

29.

30.

(c) Given that exactly one of them attended a jazz performance, what is the conditional probability that it was the younger person who attended? Suppose that a man and a woman are randomly chosen. Find the probability that (a) Exactly one attended a ballet performance. (b) At least one attended an opera. (c) Both attended a musical play. Suppose an individual is randomly chosen. Is the given table informative enough for us to determine the probability that this individual attended both a jazz and a classical music performance? If not, under what assumption would we be able to determine this probability? Compute the probability under this assumption, and then tell whether you think it is a reasonable assumption in this situation. There is a 60 percent chance that event A will occur. If A does not occur, then there is a 10 percent chance that B will occur. (a) What is the probability that at least one of the events A or B occur? (b) If A is the event that the Democratic candidate wins the presidential election in 2012 and B is the event that there is a 6.2 or higher earthquake in Los Angeles sometime in 2013, what would you take as the probability that both A and B occur? What assumption are you making? Suppose that distinct integer values are written on each of three cards. Suppose you are to be offered these cards in a random order. When you are offered a card you must immediately either accept it or reject it. If you accept a card, the process ends. If you reject a card then the next card (if a card still remains) is offered. If you reject the ﬁrst two cards offered, then you must accept the ﬁnal card. (a) If you plan to accept the ﬁrst card offered, what is the probability that you will accept the highest valued card? (b) If you plan to reject the ﬁrst card offered, and to then accept the second card if and only if its value is greater than the value of the ﬁrst card, what is the probability that you will accept the highest valued card? Let A, B, C be events such that P(A) = .2, P(B) = .3, P(C) = .4. Find the probability that at least one of the events A and B occur if (a) A and B are mutually exclusive. (b) A and B are independent. Find the probability that all of the events A, B, C occur if (c) A, B, C are independent. (d) A, B, C are mutually exclusive. Two percent of women of age 45 who participate in routine screening have breast cancer. Ninety percent of those with breast cancer have

207

208

CH A P T E R 4: Probability

positive mammographies. Ten percent of the women who do not have breast cancer will also have positive mammographies. Given a woman has a positive mammography, what is the probability she has breast cancer? 31. Identical, also called monozygotic, twins form when a single fertilized egg splits into two genetically identical parts. The twins share the same DNA set and are thus always of the same sex. Fraternal, also called dizygotic, twins develop when two separate eggs are fertilized and implant in the uterus. The genetic connection is no more or less the same as siblings born at separate times. If 64 percent of all twin pairs are of the same sex, what percentage of twin pairs are identical twins? Hint: Compute the probability that a twin pair is of the same sex by conditioning on whether the pair is monozygotic.

CHAPTER 5

Discrete Random Variables His sacred majesty, chance, decides everything. Voltaire

CONTENTS 5.1

Introduction ......................................................................... 210

5.2

Random Variables ................................................................ 211 Problems ............................................................................. 215

5.3

Expected Value .................................................................... 218 Problems ............................................................................. 225

5.4

Variance of Random Variables................................................ 231 Problems ............................................................................. 236

5.5

Binomial Random Variables ................................................... 238 Problems ............................................................................. 244

*5.6 Hypergeometric Random Variables ......................................... 248 Problems ............................................................................. 249 *5.7 Poisson Random Variables..................................................... 250 Problems ............................................................................. 253 Key Terms .................................................................................. 254 Summary .................................................................................... 254 Review Problems ......................................................................... 256

Introductory Statistics, DOI: 10.1016/B978-0-12-374388-6.00005-3 © 2010, Elsevier Inc. All rights reserved.

209

210

CH A P T E R 5: Discrete Random Variables

We continue our study of probability by introducing random variables —quantities whose values are determined by the outcome of the experiment. The expected value of a random variable is deﬁned, and its properties are explored. The concept of variance is introduced. An important special type of random variable, known as the binomial, is studied.

5.1 INTRODUCTION The National Basketball Association (NBA) draft lottery involves the 11 teams that had the worst won–lost records during the preceding year. Sixty-six Ping-Pong balls are placed in an urn. Each of these balls is inscribed with the name of a team; 11 have the name of the team with the worst record, 10 have the name of the team with the second-worst record, 9 have the name of the team with the third-worst record, and so on (with 1 ball having the name of the team with the 11th-worst record). A ball is then chosen at random, and the team whose name is on the ball is given the ﬁrst pick in the draft of players about to enter the league. All the other balls belonging to this team are then removed, and another ball is chosen. The team to which this ball “belongs” receives the second draft pick. Finally, another ball is chosen, and the team named on this ball receives the third draft pick. The remaining draft picks, 4 through 11, are then awarded to the 8 teams that did not “win the lottery,” in inverse order of their won–lost records. For instance, if the team with the worst record did not receive any of the 3 lottery picks, then that team would receive the fourth draft pick. The outcome of this draft lottery is the order in which the 11 teams get to select players. However, rather than being concerned mainly about the actual outcome, we are sometimes more interested in the values of certain speciﬁed quantities. For instance, we may be primarily interested in ﬁnding out which team gets the ﬁrst choice or in learning the draft number of our home team. These quantities of interest are known as random variables, and a special type, called discrete, will be studied in this chapter. Random variables are introduced in Sec. 5.2. In Sec. 5.3 we consider the notion of the expected value of a random variable. We see that this represents, in a sense made precise, the average value of the random variable. Properties of the expected value are presented in Sec. 5.3. Section 5.4 is concerned with the variance of a random variable, which is a measure of the amount by which a random variable tends to differ from its expected value. The concept of independent random variables is introduced in this section. Section 5.5 deals with a very important type of discrete random variable that is called binomial. We see how such random variables arise and study their properties.

5.2 Random Variables

Sections 5.6 and 5.7 introduce the hypergeometric and the Poisson random variable. We explain how these discrete random variables arise and study their properties. The ﬁrst ball drawn in the 1993 NBA draft lottery belonged to the Orlando Magic, even though the Magic had ﬁnished the season with the 11th-worst record and so had only 1 of the 66 balls!

5.2 RANDOM VARIABLES When a probability experiment is performed, often we are not interested in all the details of the experimental result, but rather are interested in the value of some numerical quantity determined by the result. For instance, in tossing dice, often we care about only their sum and are not concerned about the values on the individual dice. Also, an investor might not be interested in all the changes in the price of a stock on a given day, but rather might care about only the price at the end of the day. These quantities of interest that are determined by the result of the experiment are known as random variables. Since the value of a random variable is determined by the outcome of the experiment, we may assign probabilities to its possible values.

■

Example 5.1 The outcome of the NBA draft lottery experiment, which was discussed in Sec. 5.1, is the speciﬁcation of the teams that are to receive the ﬁrst, second, and third picks in the draft. For instance, outcome (3, 1, 4) could mean that the team with the third-worst record received pick number 1, the team with the worst record received pick number 2, and the team with the fourth-worst record received pick number 3. If we let X denote the team that received draft pick 1, then X would equal 3 if the outcome of the experiment were (3, 1, 4). Clearly, X can take on any integral value between 1 and 11 inclusive. It will equal 1 if the ﬁrst ball chosen is one of the 11 balls that belong to the team with the worst record, it will equal 2 if the ﬁrst ball is one of the 10 balls that belong to the team with the second-worst record, and so on. Since each of the 66 balls is equally likely to be the ﬁrst ball chosen, it follows that 11 66 10 P{X = 2} = 66

P{X = 1} =

5 66 4 P{X = 8} = 66 P{X = 7} =

211

212

CH A P T E R 5: Discrete Random Variables

9 66 8 P{X = 4} = 66 7 P{X = 5} = 66 6 P{X = 6} = 66 P{X = 3} =

■

3 66 2 P{X = 10} = 66 1 P{X = 11} = 66 P{X = 9} =

■

Example 5.2 Suppose we are about to learn the sexes of the three children of a certain family. The sample space of this experiment consists of the following 8 outcomes: {(b, b, b), (b, b, g), (b, g, b), (b, g, g), (g, b, b), (g, b, g), (g, g, b), (g, g, g)} The outcome (g, b, b) means, for instance, that the youngest child is a girl, the next youngest is a boy, and the oldest is a boy. Suppose that each of these 8 possible outcomes is equally likely, and so each has probability 1/8. If we let X denote the number of female children in this family, then the value of X is determined by the outcome of the experiment. That is, X is a random variable whose value will be 0, 1, 2, or 3. We now determine the probabilities that X will equal each of these four values. Since X will equal 0 if the outcome is (b, b, b), we see that P{X = 0} = P{(b, b, b)} =

1 8

Since X will equal 1 if the outcome is (b, b, g) or (b, g, b) or (g, b, b), we have P{X = 1} = P({(b, b, g), (b, g, b), (g, b, b)}) =

3 8

P{X = 2} = P({(b, g, g), (g, b, g), (g, g, b)}) =

3 8

Similarly,

P{X = 3} = P({(g, g, g)}) =

1 8

■

A random variable is said to be discrete if its possible values constitute a sequence of separated points on the number line. Thus, for instance, any random variable that can take on only a ﬁnite number of different values is discrete.

5.2 Random Variables

In this chapter we will study discrete random variables. Let X be such a quantity, and suppose that it has n possible values, which we will label x1 , x2 , . . . , xn . As in Examples 5.1 and 5.2, we will use the notation P{X = xi } to represent the probability that X is equal to xi . The collection of these probabilities is called the probability distribution of X. Since X must take on one of these n values, we know that n P{X = xi } = 1 i=1

■

Example 5.3 Suppose that X is a random variable that takes on one of the values 1, 2, or 3. If P{X = 1} = 0.4 and

P{X = 2} = 0.1

what is P{X = 3}? Solution Since the probabilities must sum to 1, we have 1 = P{X = 1} + P{X = 2} + P{X = 3} or 1 = 0.4 + 0.1 + P{X = 3} Therefore, P{X = 3} = 1 − 0.5 = 0.5 A graph of P{X = i} is shown in Fig. 5.1.

FIGURE 5.1 A graph of P{X = i}.

■

213

214

CH A P T E R 5: Discrete Random Variables

■

Example 5.4 A saleswoman has scheduled two appointments to sell encyclopedias. She feels that her ﬁrst appointment will lead to a sale with probability 0.3. She also feels that the second will lead to a sale with probability 0.6 and that the results from the two appointments are independent. What is the probability distribution of X, the number of sales made? Solution The random variable X can take on any of the values 0, 1, or 2. It will equal 0 if neither appointment leads to a sale, and so P{X = 0} = P{no sale on ﬁrst, no sale on second} = P{no sale on ﬁrst}P{no sale on second}

by independence

= (1 − 0.3)(1 − 0.6) = 0.28 The random variable X will equal 1 either if there is a sale on the ﬁrst and not on the second appointment or if there is no sale on the ﬁrst and one sale on the second appointment. Since these two events are disjoint, we have P{X = 1} = P{sale on ﬁrst, no sale on second} + P{no sale on ﬁrst, sale on second} = P{sale on ﬁrst}P{no sale on second} + P{no sale on ﬁrst}P{sale on second} = 0.3(1 − 0.6) + (1 − 0.3)0.6 = 0.54 Finally the random variable X will equal 2 if both appointments result in sales; thus P{X = 2} = P{sale on ﬁrst, sale on second} = P{sale on ﬁrst}P{sale on second} = (0.3) (0.6) = 0.18 As a check on this result, we note that P{X = 0} + P{X = 1} + P{X = 2} = 0.28 + 0.54 + 0.18 = 1

■

5.2 Random Variables

PROBLEMS 1. In Example 5.2, let the random variable Y equal 1 if the family has at least one child of each sex, and let it equal 0 otherwise. Find P{Y = 0} and P{Y = 1}. 2. In Example 5.2, let the random variable W equal the number of girls that came before the ﬁrst boy. (If the outcome is (g, g, g), take W equal to 3.) Give the possible values of W along with their probabilities. That is, give the probability distribution of W. 3. The following table presents the total number of tornadoes (violent, rotating columns of air with wind speeds over 100 miles per hour) in the United States between 1980 and 1991. Year 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 Tornadoes 866 783 1046 931 907 684 764 656 702 856 1133 1132 Source: U.S. National Oceanic and Atmospheric Administration.

Suppose that one of these years is randomly selected, and let X denote the number of tornadoes in that year. Find (a) P{X > 900} (b) P{X ≤ 800} (c) P{X = 852} (d) P{700 < X < 850} 4. Suppose a pair of dice is rolled. Let X denote their sum. What are the possible values of X? Assuming that each of the 36 possible outcomes of the experiment is equally likely, what is the probability distribution of X? 5. In Prob. 4, let Y denote the smaller of the two numbers appearing on the two dice. (If both dice show the same number, take that as the value of Y .) Determine the probability distribution of Y . 6. Two people are to meet in the park. Each person is equally likely to arrive, independent of the other, at 2:00, 2:30, or 3:00 p.m. Let X equal the time that the ﬁrst person to arrive has to wait, where X is taken to equal 0 if both people arrive at the same time. (a) What are the possible values of X? (b) What are the probabilities that X assumes each of these values? 7. Two volleyball teams are to play a 2-out-of-3 series, in which they continue to play until one has won 2 games. Suppose that the home team wins each game played, independently, with probability 0.7. Let X denote the number of games played.

215

216

CH A P T E R 5: Discrete Random Variables

8.

9.

10.

11.

(a) What are the possible values of X? (b) What is the probability distribution of X? Suppose that 2 batteries are randomly chosen from a bin containing 10 batteries, of which 7 are good and 3 are defective. Let X denote the number of defective batteries chosen. Give the possible values of X along with their probabilities. A shipment of parts contains 120 items of which 10 are defective. Two of these items are randomly chosen and inspected. Let X denote the number that are defective. Find the probability distribution of X. A contractor will bid for two jobs in sequence. She has a 0.5 probability of winning the ﬁrst job. If she wins the ﬁrst job, then she has a 0.2 chance of winning the second job; if she loses the ﬁrst job, then she has a 0.4 chance of winning the second job. (In the latter case, her bid will be lower.) Let X denote the number of jobs that she wins. Find the probability distribution of X. Whenever a certain college basketball player goes to the foul line for two shots, he makes his ﬁrst shot with probability 0.75. If he makes the ﬁrst shot, then he makes the second shot with probability 0.80; if he misses the ﬁrst shot, then he makes the second one with probability 0.70. Let X denote the number of shots he makes when he goes to the foul line for two shots. Find the probability distribution of X. In Probs. 12, 13, and 14, tell whether the set of numbers p(i), i = 1, 2, 3, 4, 5, can represent the probabilities P{X = i} of a random variable whose set of values is 1, 2, 3, 4, or 5. If your answer is no, explain why.

12.

13.

i

p (i)

1

0.4

2

0.1

3

0.2

4

0.1

5

0.3

i

p (i)

1

0.2

2

0.3

3

0.4

4 −0.1 5

0.2

5.2 Random Variables

14.

i

p (i)

1 2 3 4 5

0.3 0.1 0.2 0.4 0.0

15. In a study of 223 households in a small rural town in Iowa, a sociologist has collected data about the number of children in each household. The data showed that there are 348 children in the town, with the breakdown of the number of children in each household as follows: 38 households have 0 children, 82 have 1 child, 57 have 2 children, 34 have 3 children, 10 have 4 children, and 2 have 5 children. Suppose that one of these households is randomly selected for a more detailed interview. Let X denote the number of children in the household selected. Give the probability distribution of X. 16. Suppose that, in Prob. 15, one of the 348 children of the town is randomly selected. Let Y denote the number of children in the family of the selected child. Find the probability distribution of Y . 17. Suppose that X takes on one of the values 1, 2, 3, 4, or 5. If P{X < 3} = 0.4 and P{X > 3} = 0.5, ﬁnd (a) P{X = 3} (b) P{X < 4} 18. An insurance agent has two clients, each of whom has a life insurance policy that pays $100,000 upon death. Their probabilities of dying this year are 0.05 and 0.10. Let X denote the total amount of money that will be paid this year to the clients’ beneﬁciaries. Assuming that the event that client 1 dies is independent of the event that client 2 dies, determine the probability distribution of X. 19. A bakery has 3 special cakes at the beginning of the day. The daily demand for this type of cake is 0 1 2 3 4 5 or more

with probability 0.15 with probability 0.20 with probability 0.35 with probability 0.15 with probability 0.10 with probability 0.05

Let X denote the number of cakes that remain unsold at the end of the day. Determine the probability distribution of X.

217

218

CH A P T E R 5: Discrete Random Variables

5.3 EXPECTED VALUE A key concept in probability is the expected value of a random variable. If X is a discrete random variable that takes on one of the possible values x1 , x2 , . . . , xn , then the expected value of X, denoted by E[X], is deﬁned by E[X] =

n

xi P{X = xi }

i=1

The expected value of X is a weighted average of the possible values of X, with each value weighted by the probability that X assumes it. For instance, suppose X is equally likely to be either 0 or 1, and so P{X = 0} = P{X = 1} = then E[X] = 0

1

2

+1

1

2

=

1 2 1 2

is equal to the ordinary average of the two possible values 0 and 1 that X can assume. On the other hand, if P{X = 0} =

2 3

then

and 2

P{X = 1} = 1

1 3

1 3 3 3 is a weighted average of the two possible values 0 and 1, where the value 0 is given twice as much weight as the value 1, since it is twice as likely that X will equal 0 as it is that X will equal 1. E[X] = 0

+1

=

Deﬁnition and Terminology The expected value of a discrete random variable X whose possible values are x1 , x2 , . . . , xn , is denoted by E[X] and is deﬁned by E[X] =

n

xi P{X = xi }

i=1

Other names used for E[X] are the expectation of X and the mean of X. Another motivation for the deﬁnition of the expected value relies on the frequency interpretation of probabilities. This interpretation assumes that if a very large number (in theory, an inﬁnite number) of independent replications of an experiment are performed, then the proportion of time that event A occurs will equal P(A). Now consider a random variable X that takes on one of the possible values

5.3 Expected Value

x1 , x2 , . . . , xn , with respective probabilities p(x1 ), p(x2 ), . . . , p(xn ); and think of X as representing our winnings in a single game of chance. We will now argue that if we play a large number of such games, then our average winning per game will be E[X]. To see this, suppose that we play N games, where N is a very large number. Since, by the frequency interpretation of probability, the proportion of games in which we win xi will approximately equal p(xi ), it follows that we will win xi in approximately Np(xi ) of the N games. Since this is true for each xi , it follows that our total winnings in the N games will be approximately equal to n

xi (number of games we win xi ) =

i=1

n

xi Np(xi )

i=1

Therefore, our average winning per game will be n

i=1 xi Np(xi )

N

=

n

xi p(xi ) = E[X]

i=1

In other words, if X is a random variable associated with some experiment, then the average value of X over a large number of replications of the experiment is approximately E[X]. ■

Example 5.5 Suppose we roll a die that is equally likely to have any of its 6 sides appear face up. Find E[X], where X is the side facing up. Solution Since P[X = i] =

1 6

for i = 1, 2, 3, 4, 5, 6

we see that 1 1 1 1 1 1 E[X] = 1 +2 +3 +4 +5 +6 6 6 6 6 6 6 21 = = 3.5 6 Note that the expected value of X is not one of the possible values of X. Even though we call E[X] the expected value of X, it should be interpreted not as the value that we expect X to have, but rather as the average value of X in a large number of repetitions of the experiment. That is, if we continually roll a die, then after a large number of rolls the average of all the outcomes will be approximately 3.5. ■

219

220

CH A P T E R 5: Discrete Random Variables

■

Example 5.6 Consider a random variable X that takes on either the value 1 or 0 with respective probabilities p and 1 − p. That is, P[X = 1] = p

and

P{X = 0} = 1 − p

Find E[X]. Solution The expected value of this random variable is E[X] = 1(p) + 0(1 − p) = p ■

■

Example 5.7 An insurance company sets its annual premium on its life insurance policies so that it makes an expected proﬁt of 1 percent of the amount it would have to pay out upon death. Find the annual premium on a $200,000 life insurance policy for an individual who will die during the year with probability 0.02. Solution In units of $1000, the insurance company will set its premium so that its expected proﬁt is 1 percent of 200, or 2. If we let A denote the annual premium, then the proﬁt of the insurance company will be either A

if policyholder lives

or A − 200 if policyholder dies Therefore, the expected proﬁt is given by E[proﬁt] = AP{policyholder lives} + (A − 200)P{policyholder dies} = A(1 − 0.02) + (A − 200)(0.02) = A − 200(0.02) =A−4 So the company will have an expected proﬁt of $2000 if it charges an annual premium of $6000. ■ As seen in Example 5.7, E[X] is always measured in the same units (dollars in that example) as the random variable X.

5.3 Expected Value

FIGURE 5.2 Center of gravity = E[X].

The concept of expected value is analogous to the physical concept of the center of gravity of a distribution of mass. Consider a discrete random variable with probabilities given by p(xi ), i ≥ 1. If we imagine a rod on which weights having masses p(xi ) are placed at points xi , i ≥ 1 (Fig. 5.2), then the point at which the rod would be in balance is known as the center of gravity. It can be shown by the laws of mechanics that this point is

xi p(xi ) = E[X]

i

5.3.1 Properties of Expected Values Let X be a random variable with expected value E[X]. If c is a constant, then the quantities cX and X + c are also random variables and so have expected values. The following useful results can be shown: E[cX] = cE[X] E[X + c] = E[X] + c That is, the expected value of a constant times a random variable is equal to the constant times the expected value of the random variable; and the expected value of a constant plus a random variable is equal to the constant plus the expected value of the random variable.

■

Example 5.8 A married couple works for the same employer. The wife’s Christmas bonus is a random variable whose expected value is $1500. (a) If the husband’s bonus is set to equal 80 percent of his wife’s, ﬁnd the expected value of the husband’s bonus. (b) If the husband’s bonus is set to equal $1000 more than his wife’s, ﬁnd its expected value.

221

222

CH A P T E R 5: Discrete Random Variables

Solution Let X denote the bonus (in dollars) to be paid to the wife. (a) Since the bonus paid to the husband is equal to 0.8X, we have E[bonus to husband] = E[0.8X] = 0.8E[X] = $1200 (b) In this case the bonus to be paid to the husband is X + 1000, and so E[bonus to husband] = E[X + 1000] = E[X] + 1000 = $2500

■

A very useful property is that the expected value of the sum of random variables is equal to the sum of the individual expected values. For any random variables X and Y , E[X + Y ] = E[X] + E[Y ]

■

Example 5.9 The following are the annual incomes of 7 men and 7 women who are residents of a certain community. Annual Income (in $1000) Men

Women

33.5 25.0 28.6 41.0 30.5 29.6 32.8

24.2 19.5 27.4 28.6 32.2 22.4 21.6

Suppose that a woman and a man are randomly chosen. Find the expected value of the sum of their incomes. Solution Let X be the man’s income and Y the woman’s income. Since X is equally likely to be any of the 7 values in the men’s column, we see that

5.3 Expected Value

1 (33.5 + 25 + 28.6 + 41 + 30.5 + 29.6 + 32.8) 7 221 ≈ 31.571 = 7

E[X] =

Similarly, 1 (24.2 + 19.5 + 27.4 + 28.6 + 32.2 + 22.4 + 21.6) 7 175.9 ≈ 25.129 = 7

E[Y ] =

Therefore, the expected value of the sum of their incomes is E[X + Y ] = E[X] + E[Y ] ≈ 56.700 That is, the expected value of the sum of their incomes is approximately $56,700. ■ ■

Example 5.10 The following table lists the number of civilian full-time law enforcement employees in eight cities in 1990. City Minneapolis, MN Newark, NJ Omaha, NE Portland, OR San Antonio, TX San Jose, CA Tucson, AZ Tulsa, OK

Civilian law enforcement employees 105 155 149 195 290 357 246 178

Source: Department of Justice, Uniform Crime Reports for the United States, 1990.

Suppose that two of the cities are to be randomly chosen and all the civilian law enforcement employees of these cities are to be interviewed. Find the expected number of people who will be interviewed. Solution Let X be the number of civilian employees in the ﬁrst city chosen, and let Y be the number in the second city chosen. Since the selection of the cities is

223

224

CH A P T E R 5: Discrete Random Variables

random, each of the 8 cities has the same chance to be the ﬁrst city selected; similarly, each of the 8 cities has the same chance to be the second selection. Therefore, both X and Y are equally likely to be any of the 8 values in the given table, and so E[X] = E[Y ] = =

1 (105 + 155 + 149 + 195 + 290 + 357 + 246 + 178) 8

1675 8

and so E[X + Y ] = E[X] + E[Y ] =

1675 = 418.75 4

That is, the expected number of interviews that will be needed is 418.75.

■

By using the frequency interpretation of expected value as being the average value of a random variable over a large number of replications of the experiment, it is easy to see intuitively why the expected value of a sum is equal to the sum of the expected values. For instance, suppose we always make the same two bets on each spin of a roulette wheel, one bet concerning the color of the slot where the ball lands and the other concerning the number on that slot. Let X and Y be the amounts (in dollars) that we lose on the color bet and on the number bet, respectively, in a single spin of the wheel. Then, X + Y is our total loss in a single spin. Now, if in the long run we lose an average of 1 per spin on the color bet (so E[X] = 1) and we lose an average of 2 per spin on the number bet (so E[Y ] = 2), then our average total loss per spin (equal to E[X + Y ]) will clearly be 1 + 2 = 3. The result that the expected value of the sum of random variables is equal to the sum of the expected values holds for not only two but any number of random variables. Useful Result For any positive integer k and random variables X1 , . . . , Xk , ⎤ ⎡ k k Xi ⎦ = E[Xi ] E⎣ i=1

■

i=1

Example 5.11 A building contractor has sent in bids for three jobs. If the contractor obtains these jobs, they will yield respective proﬁts of 20, 25, and 40 (in units of $1000). On the other hand, for each job the contractor does not win, he will incur a loss (due to time and money already spent in making the bid) of 2. If the

5.3 Expected Value

probabilities that the contractor will get these jobs are, respectively, 0.3, 0.6, and 0.2, what is the expected total proﬁt? Solution Let Xi denote the proﬁt from job i, i = 1, 2, 3. Now by interpreting a loss as a negative proﬁt, we have P{X1 = 20} = 0.3

P{X1 = −2} = 1 − 0.3 = 0.7

Therefore, E[X1 ] = 20(0.3) − 2(0.7) = 4.6 Similarly, E[X2 ] = 25(0.6) − 2(0.4) = 14.2 and E[X3 ] = 40(0.2) − 2(0.8) = 6.4 The total proﬁt is X1 + X2 + X3 , and so E[total proﬁt] = E[X1 + X2 + X3 ] = E[X1 ] + E[X2 ] + E[X3 ] = 4.6 + 14.2 + 6.4 = 25.2 Therefore, the expected total proﬁt is $25,200.

PROBLEMS In the following problems, p(i) stands for P{X = i}. 1. Find the expected value of X when (a) p(1) = 1/3, p(2) = 1/3, p(3) = 1/3 (b) p(1) = 1/2, p(2) = 1/3, p(3) = 1/6 (c) p(1) = 1/6, p(2) = 1/3, p(3) = 1/2 2. Find E[X] when (a) p(1) = 0.1, p(2) = 0.3, p(3) = 0.3, p(4) = 0.2, p(5) = 0.1 (b) p(1) = 0.3, p(2) = 0.1, p(3) = 0.2, p(4) = 0.1, p(5) = 0.3 (c) p(1) = 0.2, p(2) = 0, p(3) = 0.6, p(4) = 0, p(5) = 0.2 (d) p(3) = 1

■

225

226

CH A P T E R 5: Discrete Random Variables

3. A distributor makes a proﬁt of $30 on each item that is received in perfect condition and suffers a loss of $6 on each item that is received in less-than-perfect condition. If each item received is in perfect condition with probability 0.4, what is the distributor’s expected proﬁt per item? 4. In a certain liability suit, a lawyer has to decide whether to charge a straight fee of $1200 or to take the case on a contingency basis, in which case she will receive a fee of $5000 only if her client wins the case. Determine whether the straight fee or the contingency arrangement will result in a higher expected fee when the probability that the client will win the case is (a) 1/2 (b) 1/3 (c) 1/4 (d) 1/5 5. Suppose X can take on any of the values 1, 2, and 3. Find E[X] if p(1) = 0.3

and

p(2) = 0.5

6. Let X be a random variable that is equally likely to take on any of the values 1, 2, . . . , n. That is, P{X = i} = (a) (b) (c) (d) (e)

1 n

i = 1, . . . , n

If n = 2, ﬁnd E[X]. If n = 3, ﬁnd E[X]. If n = 4, ﬁnd E[X]. For general n, what is the value of E[X]? Verify your answer in part (d) by making use of the algebraic identity n i=1

i=

n(n + 1) 2

7. A pair of fair dice is rolled. Find the expected value of the (a) Smaller (b) Larger of the two upturned faces. (If both dice show the same number, then take this to be the value of both the smaller and the larger of the upturned faces.) 8. A computer software ﬁrm has been told by its local utility company that there is a 25 percent chance that the electricity will be shut off at some time during the next working day. The company estimates

5.3 Expected Value

that it will cost $400 in lost revenues if employees do not use their computers tomorrow, and it will cost $1200 if the employees suffer a cutoff in power while using them. If the company wants to minimize the expected value of its loss, should it risk using the computers? 9. An engineering ﬁrm must decide whether to prepare a bid for a construction project. It will cost $800 to prepare a bid. If it does prepare a bid, then the ﬁrm will make a gross proﬁt (excluding the preparation cost) of $0 if it does not get the contract, $3000 if it gets the contract and the weather is bad, or $6000 if it gets the contract and the weather is not bad. If the probability of getting the contract is 0.4 and the probability that the weather will be bad is 0.6, what is the company’s expected net proﬁt if it prepares a bid? 10. All blood donated to a blood bank is tested before it is used. To reduce the total number of tests, the bank takes small samples of the blood of four separate donors and pools these samples. The pooled blood is analyzed. If it is deemed acceptable, then the bank stores the blood of these four people for future use. If it is deemed unacceptable, then the blood from each of the four donors is separately tested. Therefore, either one test or ﬁve tests are needed to handle the blood of four donors. Find the expected number of tests needed if each donor’s blood is independently unacceptable with probability 0.1. 11. Two people are randomly chosen from a group of 10 men and 20 women. Let X denote the number of men chosen, and let Y denote the number of women. (a) Find E[X]. (b) Find E[Y ]. (c) Find E[X + Y ]. 12. If the two teams in a World Series have the same chance of winning each game, independent of the results of previously played games, then the probabilities that the series will end in 4, 5, 6, or 7 games are, respectively, 1/8, 1/4, 5/16, and 5/16. What is the expected number of games played in such a series? 13. A company that operates a chain of hardware stores is planning to open a new store in one of two locations. If it chooses the ﬁrst location, the company thinks it will make a ﬁrst-year proﬁt of $40,000 if the store is successful and will have a ﬁrst-year loss of $10,000 if the store is unsuccessful. At the second location, the company thinks it will make a ﬁrst-year proﬁt of $60,000 if the store is successful and a ﬁrst-year loss of $25,000 if the store is unsuccessful. (a) If the probability of success is 1/2 for both locations, which location will result in a larger expected ﬁrst-year proﬁt? (b) Repeat part (a), this time assuming that the probability that the store is successful is 1/3.

227

228

CH A P T E R 5: Discrete Random Variables

14. If it rains tomorrow, you will earn $200 by doing some tutoring; if it is dry, you will earn $300 by doing construction work. If the probability of rain is 1/4, what is the expected amount that you will earn tomorrow? 15. If you have a 1/10 chance of gaining $400 and a 9/10 chance of gaining −$50 (that is, of losing $50), what is your expected gain? 16. If an investment has a 0.4 probability of making a $30,000 proﬁt and a 0.6 probability of losing $15,000, does this investment have a positive expected gain? 17. It costs $40 to test a certain component of a machine. If a defective component is installed, it costs $950 to repair the damage that results to the machine. From the point of view of minimizing the expected cost, determine whether the component should be installed without testing if it is known that its probability of being defective is (a) 0.1 (b) 0.05 (c) 0.01 (d) What would the probability of a defective component be if one were indifferent between testing and installing the component untested? 18. A fair bet is one in which the expected gain is equal to 0. If you bet 1 unit on a number in roulette, then you will gain 35 units if the number appears and will lose 1 unit if it does not. If the roulette wheel is perfectly balanced, then the probability that your number will appear is 1/38. What is the expected gain on a 1-unit bet? Is it a fair bet? 19. A school holding a rafﬂe will sell each ticket for $1. The school will give out seven prizes −1 for $100, 2 for $50, and 4 for $25. Suppose you purchase one ticket. If a total of 500 tickets is sold, what is your expected gain? (Hint: Your gain is −1 (if you do not win a prize), 24 (if you win a $25 prize), 49 (if you win a $50 prize), or 99 (if you win a $100 prize).) 20. A roulette wheel has 18 numbers colored red, 18 colored black, and 2 (zero and double zero) that are uncolored. If you bet 1 unit on the outcome red, then either you win 1 if a red number appears or you lose 1 if a red number does not appear. What is your expected gain? 21. The ﬁrst player to win 2 sets is the winner of a tennis match. Suppose that whatever happened in the previous sets, each player has probability 1/2 of winning the next set. Determine the expected number of sets played. 22. Suppose in Prob. 21 that the players are not of equal ability and that player 1 wins each set, independent of the results of earlier sets, with probability 1/3. (a) Find the expected number of sets played. (b) What is the probability that player 1 wins?

5.3 Expected Value

23. An insurance company sells a life insurance policy that pays $250,000 if the insured dies for an annual premium of $1400. If the probability that the policyholder dies in the course of the year is 0.005, what is the company’s expected annual proﬁt from that policyholder? 24. In Example 5.8, ﬁnd in both (a) and (b) the expected value of the sum of the bonuses earned by the wife and husband. 25. If E[X] = μ, what is E[X − μ]? 26. Four buses carrying 148 students from the same school arrive at a football stadium. The buses carry, respectively, 40, 33, 50, and 25 students. One of the students is randomly selected. Let X be the number of students who were on the bus carrying the selected student. One of 4 bus drivers is also randomly chosen. Let Y be the number of students who were on his or her bus. (a) Calculate E[X] and E[Y ]. (b) Can you give an intuitive reason why E[X] is larger than E[Y ]? 27. A small nursery must decide on the number of Christmas trees to stock. The trees cost $6 each and are to be sold for $20. Unsold trees are worthless. The nursery estimates that the probability distribution for the demand on trees is as follows: Amount demanded 1200 1500 1800 Probability

0.5

0.2

0.3

Determine the nursery’s expected proﬁt if it purchases (a) 1200 trees (b) 1500 trees (c) 1800 trees 28. Repeat Prob. 27, this time assuming that any unsold tree must be disposed of at a cost of $2 per tree. 29. The daily demand at a bakery for a certain cake is as follows: Daily demand Probability

0

1

2

3

4

0.15 0.25 0.30 0.15 0.15

It costs the bakery $4 to bake each cake, which sells for $20. Any cakes left unsold at the end of the day are thrown away. Would the bakery have a higher expected proﬁt if it baked 2 or 3 or 4 cakes daily? 30. If E[X] = 5 and E[Y ] = 12, ﬁnd (a) E[3X + 4Y ] (b) E[2 + 5Y + X] (c) E[4 + Y ]

229

230

CH A P T E R 5: Discrete Random Variables

31. Determine the expected sum of a pair of fair dice by (a) Using the probability distribution of the sum (b) Using Example 5.5 along with the fact that the expected value of the sum of random variables is equal to the sum of their expected values 32. A husband’s year-end bonus will be 0 $1000 $2000

with probability 0.3 with probability 0.6 with probability 0.1

His wife’s bonus will be $1000 $2000

with probability 0.7 with probability 0.3

Let S be the sum of their bonuses, and ﬁnd E[S]. 33. The following data give the numbers of U.S. bank failures in the years 1995 to 2002. Year

Closed or assisted

1995

8

1996

6

1997

1

1998

3

1999

8

2000

7

2001

4

2002

11

Suppose that a congressional committee has decided to randomly choose 2 of these years and then document each of the incidents that occurred in either year. Determine the expected number of such incidents. 34. Repeat Prob. 33, this time supposing that the committee is to randomly choose 3 of the years. 35. A small taxi company has 4 taxis. In a month’s time, each taxi will get 0 trafﬁc tickets with probability 0.3, 1 trafﬁc ticket with probability 0.5, or 2 trafﬁc tickets with probability 0.2. What is the expected number of tickets per month amassed by the ﬂeet of 4 taxis? 36. Suppose that 2 batteries are randomly selected from a drawer containing 8 good and 2 defective batteries. Let W denote the number of defective batteries selected.

5.4 Variance of Random Variables

(a) Find E[W] by ﬁrst determining the probability distribution of W. Let X equal 1 if the ﬁrst battery chosen is defective, and let X equal 0 otherwise. Also let Y equal 1 if the second battery is defective and equal 0 otherwise. (b) Give an equation relating X, Y , and W. (c) Use the equation in (b) to obtain E[W].

5.4 VARIANCE OF RANDOM VARIABLES It is useful to be able to summarize the properties of a random variable by a few suitably chosen measures. One such measure is the expected value. However, while the expected value gives the weighted average of the possible values of the random variable, it does not tell us anything about the variation, or spread, of these values. For instance, consider random variables U, V , and W, whose values and probabilities are as follows: U=0

with probability 1

−1 1 −10 W= 10 V=

with probability 1/2 with probability 1/2 with probability 1/2 with probability 1/2

Whereas all three random variables have expected value 0, there is clearly less spread in the values of U than in V and less spread in the values of V than in W. Since we expect a random variable X to take on values around its mean E[X], a reasonable way of measuring the variation of X is to consider how far X tends to be from its mean on the average. That is, we could consider E[|X − μ|], where μ = E[X] and |X − μ| is the absolute value of the difference between X and μ. However, it turns out to be more convenient to consider not the absolute value but the square of the difference. Deﬁnition If X is a random variable with expected value μ, then the variance of X, denoted by Var(X), is deﬁned by Var(X) = E[(X − μ)2 ] Upon expanding (X − μ)2 to obtain X 2 − 2μX + μ2 and then taking the expected value of each term, we obtain after a little algebra the following useful computational formula for Var(X): Var(X) = E[X 2 ] − μ2

(5.1)

231

232

CH A P T E R 5: Discrete Random Variables

where μ = E[X] Using Eq. (5.1) is usually the easiest way to compute the variance of X.

■

Example 5.12 Find Var(X) when the random variable X is such that X=

1 with probability p 0 with probability 1 − p

Solution In Example 5.6 we showed that E[X] = p. Therefore, using the computational formula for the variance, we have Var(X) = E[X 2 ] − p2 Now, X2 =

12 02

if X = 1 if X = 0

Since 12 = 1 and 02 = 0, we see that E[X 2 ] = 1 · P{X = 1} + 0 · P{X = 0} =1·p=p Hence, Var(X) = p − p2 = p(1 − p)

■

■

Example 5.13 The return from a certain investment (in units of $1000) is a random variable X with probability distribution P{X = −1} = 0.7

P{X = 4} = 0.2

Find Var(X), the variance of the return.

P{X = 8} = 0.1

5.4 Variance of Random Variables

Solution Let us ﬁrst compute the expected return as follows: μ = E[X] = −1(0.7) + 4(0.2) + 8(0.1) = 0.9 That is, the expected return is $900. To compute Var(X), we use the formula Var(X) = E[X 2 ] − μ2 Now, since X 2 will equal (−1)2 , 42 , or 82 with respective probabilities of 0.7, 0.2, and 0.1, we have E[X 2 ] = 1(0.7) + 16(0.2) + 64(0.1) = 10.3 Therefore, Var(X) = 10.3 − (0.9)2 = 9.49

■

5.4.1 Properties of Variances For any random variable X and constant c, it can be shown that Var(cX) = c 2 Var(X) Var(X + c) = Var(X) That is, the variance of the product of a constant and a random variable is equal to the constant squared times the variance of the random variable; and the variance of the sum of a constant and a random variable is equal to the variance of the random variable. Whereas the expected value of the sum of random variables is always equal to the sum of the expectations, the corresponding result for variances is generally not true. For instance, consider the following. Var(X + X) = Var(2X) = 22 Var(X)

= Var(X) + Var(X) However, there is an important case in which the variance of the sum of random variables is equal to the sum of the variances, and this occurs when the random

233

234

CH A P T E R 5: Discrete Random Variables

variables are independent. Before presenting this result, we must introduce the concept of independent random variables. We say that X and Y are independent if knowing the value of one of them does not change the probabilities of the other. That is, if X takes on one of the values xi , i ≥ 1, and Y takes on one of the values yj , j ≥ 1, then X and Y are independent if the events that X is equal to xi and Y is equal to yj are independent events for all xi and yj . Deﬁnition Random variables X and Y are independent if knowing the value of one of them does not change the probabilities of the other. It turns out that the variance of the sum of independent random variables is equal to the sum of their variances. Useful Result If X and Y are independent random variables, then Var(X + Y ) = Var(X) + Var(Y ) More generally, if X1 , X2 , . . . , Xk are independent random variables, then ⎛ Var ⎝

k i=1

■

⎞ Xi ⎠ =

k

Var(Xi )

i=1

Example 5.14 Determine the variance of the sum obtained when a pair of fair dice is rolled. Solution Number the dice, and let X be the value of the ﬁrst die and Y the value of the second die. Then the desired quantity is Var(X + Y ). Since the outcomes of the two dice are independent, we know that Var(X + Y ) = Var(X) + Var(Y ) To compute Var(X), the variance of the face of the ﬁrst die, recall that it was shown in Example 5.5 that E[X] =

7 2

5.4 Variance of Random Variables

Since X 2 is equally likely to be any of the values 1 2 , 22 , 32 , 42 , 52 , and 62 , we have E[X 2 ] =

1 91 (1 + 4 + 9 + 16 + 25 + 36) = 6 6

Therefore, 2 7 2 91 49 − = 6 4 35 = 12

Var(X) = E[X 2 ] −

Since Y has the same probability distribution as X, it also has variance 35/12, and so Var(X + Y ) =

35 35 35 + = 12 12 6

■

The positive square root of the variance is called the standard deviation (SD). Deﬁnition The quantity SD(X), deﬁned by SD(X) = Var(X) is called the standard deviation of X. The standard deviation, like the expected value, is measured in the same units as is the random variable. That is, if the value of X is given in terms of miles, then so will the expected value and the standard deviation, too. To compute the standard deviation of a random variable, compute the variance and then take its square root. ■

Example 5.15 The annual gross earnings of a certain rock singer are a random variable with an expected value of $400,000 and a standard deviation of $80,000. The singer’s manager receives 15 percent of this amount. Determine the expected value and standard deviation of the amount received by the manager. Solution If we let X denote the earnings (in units of $1000) of the singer, then the manager earns 0.15X. Its expected value is obtained as follows: E[0.15X] = 0.15E[X] = 60

235

236

CH A P T E R 5: Discrete Random Variables

To compute the standard deviation, ﬁrst determine the variance: Var(0.15X) = (0.15)2 Var(X) Taking the square root of both sides of the preceding gives SD(0.15X) = 0.15 SD(X) = 12 Therefore, the amount received by the manager is a random variable with an expected value of $60,000 and a standard deviation of $12,000. ■

PROBLEMS 1. Determine the variances of random variables U, V , and W, deﬁned at the beginning of Sec. 5.4. 2. Let p(i) = P{X = i}. Consider (a) p(0) = 0.50, p(1) = 0.50 (b) p(0) = 0.60, p(1) = 0.40 (c) p(0) = 0.90, p(1) = 0.10 In which case do you think Var(X) would be largest? And in which case would it be smallest? Determine the actual variances and check your answers. 3. Suppose that, for some constant c, P{X = c} = 1. Find Var(X). 4. Find the variances of the random variables speciﬁed in Prob. 1 of Sec. 5.3. 5. Find Var(X) for the X given in Prob. 5 of Sec. 5.3. 6. If the probability that you earn $300 is 1/3 and the probability that you earn $600 is 2/3, what is the variance of the amount that you earn? 7. Find the variance of the number of sets played in the situation described in Prob. 21 of Sec. 5.3. 8. A small electronics company that started up 4 years ago has 60 employees. The following is a frequency table relating the number of years (rounded up) that these employees have been with the company. Number of years Frequency 1

12

2

25

3

16

4

7

5.4 Variance of Random Variables

Suppose one of these workers is randomly chosen. Let X denote the number of years he or she has been with the company. Find (a) E[X] (b) Var(X) 9. The vacation time received by a worker of a certain company depends on the economic performance of the company. Suppose that Fong, an employee of this company, will receive 0 weeks’ vacation 1 week’s vacation 2 weeks’ vacation

with probability 0.4 with probability 0.2 with probability 0.4

Suppose also that Fontanez, another employee, will receive 0 weeks’ vacation 1 week’s vacation 2 weeks’ vacation

10. 11.

12.

13.

14. 15.

with probability 0.3 with probability 0.4 with probability 0.3

Let X denote the number of weeks of vacation for Fong and Y denote the number of weeks for Fontanez. (a) Which do you think is larger, Var(X) or Var(Y )? (b) Find Var(X). (c) Find Var(Y ). Find the variance of the proﬁt earned by the nursery in Prob. 27(b) of Sec. 5.3. Two fair coins are tossed. Determine Var(X) when X is the number of heads that appear. (a) Use the deﬁnition of the variance. (b) Use the fact that the variance of the sum of independent random variables is equal to the sum of the variances. Find the variance of the number of tickets obtained by the ﬂeet of taxis, as described in Prob. 35 of Sec. 5.3. Assume that the numbers of tickets received by each of the taxis are independent. A lawyer must decide whether to charge a ﬁxed fee of $2000 or to take a contingency fee of $8000 if she wins the case (and $0 if she loses). She estimates that her probability of winning is 0.3. Determine the standard deviation of her fee if (a) She takes the ﬁxed fee. (b) She takes the contingency fee. Find the standard deviation of the amount of money you will earn in Prob. 14 of Sec. 5.3. The following is a frequency table giving the number of courses being taken by 210 ﬁrst-year students at a certain college.

237

238

CH A P T E R 5: Discrete Random Variables

Number of classes Frequency 1 2 3 4 5 6 7

16.

17. 18. 19.

2 15 37 90 49 14 3

Let X denote the number of courses taken by a randomly chosen student. Find (a) E[X] (b) SD(X) The amount of money that Robert earns has expected value $30,000 and standard deviation $3000. The amount of money that his wife Sandra earns has expected value $32,000 and standard deviation $5000. Determine the (a) Expected value (b) Standard deviation of the total earnings of this family. In answering part (b), assume that Robert’s earnings and Sandra’s earnings are independent. (Hint: In answering part (b), ﬁrst ﬁnd the variance of the family’s total earnings.) If Var(X) = 4, what is SD(3X)? (Hint: First ﬁnd Var(3X).) If Var(2X + 3) = 16, what is SD(X)? If X and Y are independent random variables, both having variance 1, ﬁnd (a) Var(X + Y ) (b) Var(X − Y )

5.5 BINOMIAL RANDOM VARIABLES One of the most important types of random variables is the binomial, which arises as follows. Suppose that n independent subexperiments (or trials) are performed, each of which results in either a “success” with probability p or a “failure” with probability 1 − p. If X is the total number of successes that occur in n trials, then X is said to be a binomial random variable with parameters n and p. Before presenting the general formula for the probability that a binomial random variable X takes on each of its possible values 0, 1, . . . , n, we consider a special case. Suppose that n = 3 and that we are interested in the probability that X is

5.5 Binomial Random Variables

equal to 2. That is, we are interested in the probability that 3 independent trials, each of which is a success with probability p, will result in a total of 2 successes. To determine this probability, consider all the outcomes that give rise to exactly 2 successes: (s, s, f ), (s, f , s), (f , s, s) The outcome (s, f , s) means, for instance, that the ﬁrst trial is a success, the second a failure, and the third a success. Now, by the assumed independence of the trials, it follows that each of these outcomes has probability p2 (1 − p). For instance, if Si is the event that trial i is a success and Fi is the event that trial i is a failure, then P(s, f , s) = P(S1 ∩ F2 ∩ S3 ) = P(S1 )P(F2 )P(S3 )

by independence

= p(1 − p)p Since each of the 3 outcomes that result in a total of 2 successes consists of 2 successes and 1 failure, it follows in a similar fashion that each occurs with probability p2 (1 − p). Therefore, the probability of a total of 2 successes in the 3 trials is 3p2 (1 − p). Consider now the general case in which we have n independent trials. Let X denote the number of successes. To determine P{X = i}, consider any outcome that results in a total of i successes. Since this outcome will have a total of i successes and n − i failures, it follows from the independence of the trials that its probability will be pi (1 − p)n−i . That is, each outcome that results in X = i will have the same probability pi (1 − p)n−i . Therefore, P{X = i} is equal to this common probability multiplied by the number of different outcomes that result in i successes. Now, it can be shown that there are n!/[i!(n − i)!] different outcomes that result in a total of i successes and n − i failures, where n! (read “n factorial”) is equal to 1 when n = 0 and is equal to the product of the natural numbers from 1 to n otherwise. That is, 0! = 1 n! = n · (n − 1) · · · 3 · 2 · 1 if n > 0 A binomial random variable with parameters n and p represents the number of successes in n independent trials, when each trial is a success with probability p. If X is such a random variable, then for i = 0, . . . , n, P{X = i} =

n! pi (1 − p)n−i i! (n − i)!

239

240

CH A P T E R 5: Discrete Random Variables

As a check of the preceding equation, note that it states that the probability that there are no successes in n trials is p{X = 0} =

n! p0 (1 − p)n−0 0! (n − 0)!

= (1 − p)n

since 0! = p0 = 1

However, the foregoing is clearly correct since the probability that there are 0 successes, and so all the trials are failures, is, by independence, (1 − p)(1 − p) · · · (1 − p) = (1 − p)n . The probabilities of three binomial random variables with respective parameters n = 10, p = 0.5, n = 10, p = 0.3, and n = 10, p = 0.6 are presented in Fig. 5.3. ■

Example 5.16 Three fair coins are ﬂipped. If the outcomes are independent, determine the probability that there are a total of i heads, for i = 0, 1, 2, 3. Solution If we let X denote the number of heads (“successes”), then X is a binomial random variable with parameters n = 3, p = 0.5. By the preceding we have 3 3! 1 0 1 3 1 1 = = P{X = 0} = 0! 3! 2 2 2 8 3 1 2 3! 1 1 1 3 P{X = 1} = =3 = 1! 2! 2 2 2 8 3 2 1 3! 1 1 1 3 P{X = 2} = =3 = 2! 1! 2 2 2 8 3 3! 1 3 1 0 1 1 = = P{X = 3} = 3! 0! 2 2 2 8 ■

■

Example 5.17 Suppose that a particular trait (such as eye color or handedness) is determined by a single pair of genes, and suppose that d represents a dominant gene and r a recessive gene. A person with the pair of genes (d, d) is said to be pure dominant, one with the pair (r, r) is said to be pure recessive, and one with the pair (d, r) is said to be hybrid. The pure dominant and the hybrid are alike in appearance. When two individuals mate, the resulting offspring receives one gene from each parent, and this gene is equally likely to be either of the parent’s two genes.

5.5 Binomial Random Variables

FIGURE 5.3 Binomial probabilities.

(a) What is the probability that the offspring of two hybrid parents has the opposite (recessive) appearance? (b) Suppose two hybrid parents have 4 offsprings. What is the probability 1 of the 4 offspring has the recessive appearance?

241

242

CH A P T E R 5: Discrete Random Variables

Solution (a) The offspring will have the recessive appearance if it receives a recessive gene from each parent. By independence, the probability of this is (1/2)(1/2) = 1/4. (b) Assuming the genes obtained by the different offspring are independent (which is the common assumption in genetics), it follows from part (a) that the number of offspring having the recessive appearance is a binomial random variable with parameters n = 4 and p = 1/4. Therefore, if X is the number of offspring that have the recessive appearance, then 4! 1 1 3 3 P{X = 1} = 1! 3! 4 4 3 1 3 =4 4 4 =

27 64

Suppose that X is a binomial random variable with parameters n and p, and suppose we want to calculate the probability that X is less than or equal to some value j. In principle, we could compute this as follows:

P{X ≤ j} =

j i=0

P{X = i} =

j i=0

n! pi (1 − p)n−i i! (n − i)!

The amount of computation called for in the preceding equation can be rather large. To relieve this, Table D.5 (in App. D) gives the values of P{X ≤ j} for n ≤ 20 and for various values of p. In addition, you can use Program 5-1. In this program you enter the binomial parameters and the desired value of j, and you get as output the probability that the binomial is less than or equal to j, the probability that the binomial is equal to j, and the probability that the binomial is greater than or equal to j. ■ ■

Example 5.18 (a) Determine P{X ≤ 12} when X is a binomial random variable with parameters 20 and 0.4. (b) Determine P{Y ≤ 10} when Y is a binomial random variable with parameters 16 and 0.5. Solution From Table D.5, we see that

5.5 Binomial Random Variables

(a) P{X ≤ 12} = 0.9790 (b) P{Y ≤ 10} = 1 − P{Y < 10} = 1 − P{Y ≤ 9} = 1 − 0.7728 = 0.2272 We could also have run Program 5-1 to obtain the following: The probability that a binomial (20, 0.4) is less than or equal to 12 is 0.978969. The probability that a binomial (16, 0.5) is greater than or equal to 10 is 0.2272506. ■

5.5.1 Expected Value and Variance of a Binomial Random Variable A binomial (n, p) random variable X is equal to the number of successes in n independent trials when each trial is a success with probability p. As a result, we can represent X as X=

n

Xi

i=1

where Xi is equal to 1 if trial i is a success and is equal to 0 if trial i is a failure. Since P{Xi = 1} = p

and

P{Xi = 0} = 1 − p

it follows from the results of Examples 5.6 and 5.12 that E[Xi ] = p

and Var(Xi ) = p(1 − p)

Therefore, using the fact that the expectation of the sum of random variables is equal to the sum of their expectations, we see that E[X] = np Also, since the variance of the sum of independent random variables is equal to the sum of their variances, we have Var(X) = np(1 − p) Let us summarize. If X is binomial with parameters n and p, then E[X] = np Var(X) = np(1 − p)

243

CH A P T E R 5: Discrete Random Variables

■

Example 5.19 Suppose that each screw produced is independently defective with probability 0.01. Find the expected value and variance of the number of defective screws in a shipment of size 1000. Solution The number of defective screws in the shipment of size 1000 is a binomial random variable with parameters n = 1000, p = 0.01. Hence, the expected number of defective screws is E[number of defectives] = 1000(0.01) = 10 and the variance of the number of defective screws is Var(number of defectives) = 1000(0.01)(0.99) = 9.9

■

Historical Perspective Bettmann

244

Jacques Bernoulli

Independent trials having a common success probability p were ﬁrst studied by the Swiss mathematician Jacques Bernoulli (1654–1705). In his book Ars Conjectandi (The Art of Conjecturing), published by his nephew Nicholas eight years after his death in 1713, Bernoulli showed that if the number of such trials were large, then the proportion of them that were successes would be close to p with a probability near 1. Jacques Bernoulli was from the ﬁrst generation of the most famous mathematical family of all time. Altogether there were anywhere between 8 and 12 Bernoullis, spread over three generations, who made fundamental contributions to probability, statistics, and mathematics. One difﬁculty in knowing their exact number is the fact that several had the same name. (For example, two of the sons of Jacques’ brother Jean were named Jacques and Jean.) Another difﬁculty is that several of the Bernoullis were known by different names in different places. Our Jacques (sometimes written Jaques), for instance, was also known as Jakob (sometimes written Jacob) and as James Bernoulli. But whatever their number, their inﬂuence and output were prodigious. Like the Bachs of music, the Bernoullis of mathematics were a family for the ages!

PROBLEMS 1. Find (a) 4! (b) 5! (c) 7! 8! 7! 9! (b) (c) 2. Find (a) 3! 5! 3! 4! 4! 5!

5.5 Binomial Random Variables

3. Given that 9! = 362,880, ﬁnd 10!. 4. Use the probability distribution of a binomial random variable with parameters n and p to show that P{X = n} = pn Then argue directly why this is valid. 5. If X is a binomial random variable with parameters n = 8 and p = 0.4, ﬁnd (a) P{X = 3} (b) P{X = 5} (c) P{X = 7} 6. Each ball bearing produced is independently defective with probability 0.05. If a sample of 5 is inspected, ﬁnd the probability that (a) None are defective. (b) Two or more are defective. 7. Suppose you will be attending 6 hockey games. If each game independently will go to overtime with probability 0.10, ﬁnd the probability that

8.

9.

10.

11.

(a) At least 1 of the games will go into overtime. (b) At most 1 of the games will go into overtime. A satellite system consists of 4 components and can function if at least 2 of them are working. If each component independently works with probability 0.8, what is the probability the system will function? A communications channel transmits the digits 0 and 1. Because of static, each digit transmitted is independently incorrectly received with probability 0.1. Suppose an important single-digit message is to be transmitted. To reduce the chance of error, the string of digits 0 0 0 0 0 is to be transmitted if the message is 0 and the string 1 1 1 1 1 is to be transmitted if the message is 1. The receiver of the message uses “majority rule” to decode; that is, she decodes the message as 0 if there are at least 3 zeros in the message received and as 1 otherwise. (a) For the message to be incorrectly decoded, how many of the 5 digits received would have to be incorrect? (b) What is the probability that the message is incorrectly decoded? A multiple-choice examination has 3 possible answers for each of 5 questions. What is the probability that a student will get 4 or more correct answers just by guessing? A man claims to have extrasensory perception (ESP). As a test, a fair coin is to be ﬂipped 8 times, and he is asked to predict the outcomes in advance. Suppose he gets 6 correct answers. What is the probability that he would have got at least this number of correct answers if he had no ESP but had just guessed?

245

246

CH A P T E R 5: Discrete Random Variables

12. Each diskette produced by a certain company will be defective with probability 0.05 independent of the others. The company sells the diskettes in packages of 10 and offers a money-back guarantee that all the diskettes in a package will be nondefective. Suppose that this offer is always taken up. (a) What is the probability that a package is returned? (b) If someone buys 3 packages, what is the probability that exactly 1 of them is returned? 13. Four fair dice are to be rolled. Find the probability that (a) 6 appears at least once. (b) 6 appears exactly once. (c) 6 appears at least twice. 14. Statistics indicate that alcohol is a factor in 55 percent of fatal automobile accidents. Of the next 3 fatal automobile accidents, ﬁnd the probability that alcohol is a factor in (a) All 3 (b) Exactly 2 (c) At least 1 15. Individuals who have two sickle cell genes will develop the disease called sickle cell anemia, while individuals having none or one sickle cell gene will not be harmed. If two people, both of whom have one sickle cell gene, have a child, then that child will receive two sickle cell genes with probability 1/4. Suppose that both members of each of three different couples have exactly one sickle cell gene. If each of these couples has a child, ﬁnd the probability that (a) None of the children receives two sickle cell genes. (b) Exactly one of the children receives two sickle cell genes. (c) Exactly two of the children receive two sickle cell genes. (d) All three children receive two sickle cell genes. 16. Let X be a binomial random variable with parameters n = 20 and p = 0.6. Find (a) P{X ≤ 14} (b) P{X < 10} (c) P{X ≥ 13}

(d) P{X > 10} (e) P{9 ≤ X ≤ 16} (f) P{7 < X < 15}

17. A fair die is to be rolled 20 times. Find the expected value of the number of times (a) 6 appears. (b) 5 or 6 appears.

(c) An even number appears. (d) Anything else but 6 appears.

18. Find the variances of the random variables in Prob. 17.

5.5 Binomial Random Variables

19. The probability that a ﬂuorescent bulb burns for at least 500 hours is 0.90. Of 8 such bulbs, ﬁnd the probability that (a) All 8 burn for at least 500 hours. (b) Exactly 7 burn for at least 500 hours. (c) What is the expected value of the number of bulbs that burn for at least 500 hours? (d) What is the variance of the number of bulbs that burn for at least 500 hours? 20. If a fair coin is ﬂipped 500 times, what is the standard deviation of the number of times that a head appears? 21. The FBI has reported that 44 percent of murder victims are killed with handguns. If 4 murder victims are randomly selected, ﬁnd (a) The probability that they were all killed by handguns (b) The probability that none were killed by handguns (c) The probability that at least two were killed by handguns (d) The expected number killed by handguns (e) The standard deviation of the number killed by handguns 22. The expected number of heads in a series of 10 ﬂips of a coin is 6. What is the probability there are 8 heads? 23. If X is a binomial random variable with expected value 4 and variance 2.4, ﬁnd (a) P{X = 0} (b) P{X = 12} 24. If X is a binomial random variable with expected value 4.5 and variance 0.45, ﬁnd (a) P{X = 3} (b) P{X ≥ 4} 25. Find the mean and standard deviation of a binomial random variable with parameters (a) n = 100, p = 0.5 (d) n = 50, p = 0.5 (b) n = 100, p = 0.4 (e) n = 150, p = 0.5 (c) n = 100, p = 0.6 (f) n = 200, p = 0.25 26. The National Basketball Association championship series is a best-ofseven series, meaning that the ﬁrst team to win four games is declared the champion. In its history, no team has ever come back to win the championship after being behind three games to one. Assuming that each of the games played in this year’s series is equally likely to be won by either team, independent of the results of earlier games, what is the probability that the upcoming championship series will be the ﬁrst time that a team comes back from a three-game-to-one deﬁcit to win the series?

247

248

CH A P T E R 5: Discrete Random Variables

*5.6 HYPERGEOMETRIC RANDOM VARIABLES Suppose that n batteries are to be randomly selected from a bin of N batteries, of which Np are functional and the other N(1 − p) are defective. The random variable X, equal to the number of functional batteries in the sample, is then said to be a hypergeometric random variable with parameters n, N, p. We can interpret the preceding experiment as consisting of n trials, where trial i is considered a success if the ith battery withdrawn is a functional battery. Since each of the N batteries is equally likely to be the ith one withdrawn, it follows that trial i is a success with probability Np/N = p. Therefore, X can be thought of as representing the number of successes in n trials where each trial is a success with probability p. What distinguishes X from a binomial random variable is that these trials are not independent. For instance, suppose that two batteries are to be withdrawn from a bin of ﬁve batteries, of which one is functional and the others defective. (That is, n = 2, N = 5, p = 1/5.) Then the probability that the second battery withdrawn is functional is 1/5. However, if the ﬁrst one withdrawn is functional, then the conditional probability that the second one is functional is 0 (since when the second battery is chosen all four remaining batteries in the bin are defective). That is, when the selections of the batteries are made without replacing the previously chosen ones, the trials are not independent, so X is not a binomial random variable. By using the result that each of the n trials is a success with probability p, it can be shown that the expected number of successes is np. That is, E[X] = np In addition, it can be shown that the variance of the hypergeometric random variable is given by Var(X) =

N−n np(1 − p) N−1

Thus, whereas the expected value of the hypergeometric random variable with parameters n, N, p is the same as that of the binomial random variable with parameters n, p, its variance is smaller than that of the binomial by the factor (N − n)/(N − 1). ■

Example 5.20 If 6 people are randomly selected from a group consisting of 12 men and 8 women, then the number of women chosen is a hypergeometric random variable with parameters n = 6, N = 20, p = 8/20 = 0.4. Its mean and variance are E[X] = 6(0.4) = 2.4 Var(X) =

14 6(0.4)(0.6) ≈ 1.061 19

5.6 Hypergeometric Random Variables

Similarly, the number of men chosen is a hypergeometric random variable with parameters n = 6, N = 20, p = 0.6. ■ Suppose now that N, the number of batteries in the urn, is large in comparison to n, the number to be selected. For instance, suppose that 20 batteries are to be randomly chosen from a bin containing 10,000 batteries of which 90 percent are functional. In this case, no matter which batteries were previously chosen each new selection will be defective with a probability that is approximately equal to 0.9. For instance, the ﬁrst battery selected will be functional with probability 0.9. If the ﬁrst battery is functional then the next one will also be functional with probability 8999/9999 ≈ .89999, whereas if the ﬁrst battery is defective then the second one will be functional with probability 9000/9999 ≈ .90009. A similar argument holds for the other selections, and thus we may conclude that when N is large in relation to n, then the n trials are approximately independent, which means that X is approximately a binomial random variable. When N is large in relation to n, a hypergeometric random variable with parameters n, N, p approximately has a binomial distribution with parameters n and p.

PROBLEMS In the following problems, state whether the random variable X is binomial or hypergeometric. Also give its parameters (n and p if it is binomial or n, N, and p if it is hypergeometric). 1. A lot of 200 items contains 18 defectives. Let X denote the number of defectives in a sample of 20 items. 2. A restaurant knows from past experience that 15 percent of all reservations do not show. Twenty reservations are expected tonight. Let X denote the number that show. 3. In one version of the game of lotto each player selects six of the numbers from 1 to 54. The organizers also randomly select six of these numbers. These latter six are called the winning numbers. Let X denote how many of a given player’s six selections are winning numbers. 4. Each new fuse produced is independently defective with probability 0.05. Let X denote the number of defective fuses in the last 100 produced. 5. Suppose that a collection of 100 fuses contains 5 that are defective. Let X denote the number of defectives discovered when 20 of them are randomly chosen and inspected. 6. A deck of cards is shufﬂed and the cards are successively turned over. Let X denote the number of aces in the ﬁrst 10 cards.

249

250

CH A P T E R 5: Discrete Random Variables

7. A deck of cards is shufﬂed and the top card is turned over. The card is then returned to the deck and the operation repeated. This continues until a total of 10 cards have been turned over. Let X denote the number of aces that have appeared.

*5.7 POISSON RANDOM VARIABLES A random variable X that takes on one of the values 0, 1, 2, . . . is said to be a Poisson random variable with parameter λ if for some positive value λ its probabilities are given by P{X = i} = cλi /i!,

i = 0, 1, . . .

In the preceding, c is a constant that depends on λ. Its explicit value is given by c = e−λ , where e is a famous mathematical constant that is approximately equal to 2.718. A random variable X is called a Poisson random variable with parameter λ if P{X = i} =

e−λ λi , i!

i = 0, 1, . . .

A graph of the probabilities of a Poisson random variable having parameter λ = 4 is presented in Fig. 5.4.

■

Example 5.21 If X is a Poisson random variable with parameter λ = 2, ﬁnd P{X = 0}. Solution P{X = 0} =

e−2 20 0!

Using the facts that 20 = 1 and 0! = 1, we obtain P{X = 0} = e−2 = 0.1353 In the preceding, the value of e−2 was obtained from a table of exponentials. Alternatively, it could have been obtained from a scientiﬁc hand calculator or a personal computer. ■ Poisson random variables arise as approximations to binomial random variables. Consider n independent trials, each of which results in either a success with probability p or a failure with probability 1 − p. If the number of trials is large and the

5.7 Poisson Random Variables

FIGURE 5.4 Probabilities of a Poisson random variable with λ = 4.

probability of a success on a trial is small, then the total number of successes will be approximately a Poisson random variable with parameter λ = np. Some examples of random variables whose probabilities are approximately given, for some λ, by Poisson probabilities are the following: 1. The number of misprints on a page of a book 2. The number of people in a community who are at least 100 years old 3. The number of people entering a post ofﬁce on a given day Each of these is approximately Poisson because of the Poisson approximation to the binomial. For instance, we can suppose that each letter typed on a page has a small probability of being a misprint, and so the number of misprints on a page will be approximately a Poisson random variable with parameter λ = np, where n is the large number of letters on a page and p is the small probability that any given letter is a misprint. ■

Example 5.22 Suppose that items produced by a certain machine are independently defective with probability 0.1. What is the probability that a sample of 10 items will contain at most 1 defective item? What is the Poisson approximation for this probability?

251

252

CH A P T E R 5: Discrete Random Variables

Solution If we let X denote the number of defective items, then X is a binomial random variable with parameters n = 10, p = 0.1. Thus the desired probability is P{X = 0} + P{X = 1} =

10 10 (0.1)0 (0.9)10 + (0.1)1 (0.9)9 0 1

= 0.7361 Since np = 10(0.1) = 1, the Poisson approximation yields the value P{X = 0} + P{X = 1} = e−1 + e−1 = 0.7358 Thus, even in this case, where n is equal to 10 (which is not that large) and p is equal to 0.1 (which is not that small), the Poisson approximation to the binomial probability is quite accurate. ■ Both the expected value and the variance of a Poisson random variable are equal to λ. That is, we have the following. If X is a Poisson random variable with parameter λ, λ > 0, then E[X] = λ Var(X) = λ

■

Example 5.23 Suppose the average number of accidents occurring weekly on a particular highway is equal to 1.2. Approximate the probability that there is at least one accident this week. Solution Let X denote the number of accidents. Because it is reasonable to suppose that there are a large number of cars passing along the highway, each having a small probability of being involved in an accident, the number of such accidents should be approximately a Poisson random variable. That is, if X denotes the number of accidents that will occur this week, then X is approximately a Poisson random variable with mean value λ = 1.2. The desired probability is now

5.7 Poisson Random Variables

obtained as follows: P{X > 0} = 1 − P{X = 0} =1−

e−1.2 (1.2)0 0!

= 1 − e−1.2 = 1 − 0.3012 = 0.6988 Therefore, there is approximately a 70 percent chance that there will be at least one accident this week. ■

PROBLEMS The following will be needed for the problems. The values given are correct to four decimal places. e−1/2 = 0.6065,

e−4 = 0.0183,

e−1 = 0.3679,

e−0.3 = 0.7408

1. If X is Poisson with mean λ = 4, ﬁnd (a) P{X = 1} (b) P{X = 2} (c) P{X > 2} 2. Compare the Poisson approximation with the true binomial probability in the following cases: (a) P{X = 2} when n = 10, p = 0.1 (b) P{X = 2} when n = 10, p = 0.05 (c) P{X = 2} when n = 10, p = 0.01 (d) P{X = 2} when n = 10, p = 0.3 3. You buy a lottery ticket in 500 lotteries. In each lottery your chance of winning a prize is 1/1000. What is the approximate probability for the following? (a) You win 0 prizes. (b) You win exactly 1 prize. (c) You win at least 2 prizes. 4. If X is Poisson with mean λ = 144, ﬁnd (a) E[X] (b) SD(X)

253

254

CH A P T E R 5: Discrete Random Variables

5. A particular insurance company pays out an average of 4 major medical claims in a month. (a) Approximate the probability that it pays no major medical claims in the coming month? (b) Approximate the probability that it pays at most 2 major medical claims in the coming month? (c) Approximate the probability that it pays at least 4 major medical claims in the coming month?

KEY TERMS Random variable: A quantity whose value is determined by the outcome of a probability experiment. Discrete random variable: A random variable whose possible values constitute a sequence of disjoint points on the number line. Expected value of a random variable: A weighted average of the possible values of a random variable; the weight given to a value is the probability that the random variable is equal to that value. Also called the expectation or the mean of the random variable. Variance of a random variable: The expected value of the square of the difference between the random variable and its expected value. Standard deviation of a random variable: The square root of the variance. Independent random variables: A set of random variables having the property that knowing the values of any subset of them does not affect the probabilities of the remaining ones. Binomial random variable with parameters n and p: A random variable equal to the number of successes in n independent trials when each trial is a success with probability p.

SUMMARY A random variable is a quantity whose value is determined by the outcome of a probability experiment. If its possible values can be written as a sequence of distinct numbers, then the random variable is called discrete. Let X be a random variable whose possible values are xi , i = 1, . . . , n; and suppose X takes on the value xi with probability P{X = xi }. The expected value of X, also referred to as the mean of X or the expectation of X, is denoted by E[X] and is deﬁned as n E[X] = xi P{X = xi } i=1

Summary

If X is a random variable and c is a constant, then E[cX] = cE[X] E[X + c] = E[X] + c For any random variables X1 , . . . , Xk , E[X1 + X2 + · · · + Xk ] = E[X1 ] + E[X2 ] + · · · + E[Xk ] The random variables X and Y are independent if knowing the value of one of them does not change the probabilities for the other. The variance of a random variable measures the average squared distance of the random variable from its mean. Speciﬁcally, if X has mean μ = E[X], then the variance of X, denoted by Var(X), is deﬁned as Var(X) = E[(X − μ)2 ] A property of the variance is that for any constant c and random variable X, Var(cX) = c 2 Var(X) Var(X + c) = Var(X) Whereas the variance of the sum of random variables in general is not equal to the sum of their variances, it is true in the special case where the random variables are independent. That is, Var(X + Y ) = Var(X) + Var(Y ) if X and Y are independent. The square root of the variance is called the standard deviation and is denoted by SD(X). That is, SD(X) + Var(X) Consider n independent trials in which each trial results in a success with probability p. If X is the total number of successes, then X is said to be a binomial random variable with parameters n and p. Its probabilities are given by P{X = i} =

n! pi (1 − p)n−i i!(n − i)!

i = 0, . . . , n

In the above, n! (called n factorial) is deﬁned by 0! = 1

n! = n(n − 1) . . . 3 · 2 · 1

255

256

CH A P T E R 5: Discrete Random Variables

The mean and variance of a binomial random variable with parameters n and p are E[X] = np

and Var (X) = np(1 − p)

Binomial random variables with large values of n and small values of p can be approximated by a Poisson random variable, whose probabilities are given by P{X = i} = e−λ

λi , i!

i = 0, 1, . . .

where λ = np. The mean and variance of this random variable are both equal to λ.

REVIEW PROBLEMS 1. If P{X ≤ 4} = 0.8 and P{X = 4} = 0.2, ﬁnd (a) P{X ≥ 4} (b) P{X < 4} 2. If P{X ≤ 6} = 0.7 and P{X < 6} = 0.5, ﬁnd (a) P{X = 6} (b) P{X > 6} 3. A graduating law student is not certain whether he actually wants to practice law or go into business with his family. He has decided to base his decision on whether he can pass the bar examination. He has decided to give himself at most 4 attempts at the examination; he will practice law if he passes the examination or go into the family business if he fails on all 4 tries. Suppose that each time he takes the bar examination he is successful, independent of his previous results, with probability 0.3. Let X denote the number of times he takes the bar examination. (a) What are the possible values that X can assume? (b) What is the probability distribution of X? (c) What is the probability that he passes the bar examination? (d) Find E[X]. (e) Find Var (X). 4. Suppose that X is either 1 or 2. If E[X] = 1.6, ﬁnd P{X = 1}. 5. A gambling book recommends the following “winning strategy” for the game of roulette. It recommends that a gambler bet $1 on red. If red appears (which has probability 18/38 of occurring), then the gambler should take her $1 proﬁt and quit. If the gambler loses this bet (which has probability 20/38 of occurring) and so is behind $1, then she should make a $2 bet on red and then quit. Let X denote the gambler’s ﬁnal winnings.

Review Problems

6.

7.

8.

9.

10. 11.

(a) Find P{X > 0}. (b) Are you convinced that the strategy is a winning strategy? Why or why not? (c) Find E[X]. Two people are to meet in the park. Each person is equally likely to arrive, independent of the other, at 3:00, 4:00, or 5:00 p.m. Let X equal the time that the ﬁrst person to arrive has to wait, where X is taken to equal 0 if they both arrive at the same time. Find E[X]. There is a 0.3 probability that a used-car salesman will sell a car to his next customer. If he does, then the car that is purchased is equally likely to cost $4000 or $6000. Let X denote the amount of money that the customer spends. (a) Find the probability distribution of X. (b) Find E[X]. (c) Find Var (X). (d) Find SD (X). Suppose that 2 batteries are randomly chosen from a bin containing 12 batteries, of which 8 are good and 4 are defective. What is the expected number of defective batteries chosen? A company is preparing a bid for a contract to supply a city’s schools with notebook supplies. The cost to the company of supplying the material is $140,000. It is considering two alternate bids: to bid high (25 percent above cost) or to bid low (10 percent above cost). From past experience the company knows that if it bids high, then the probability of winning the contract is 0.15, whereas if it bids low, then the probability of winning the contract is 0.40. Which bid will maximize the company’s expected proﬁt? If E[3X + 10] = 70, what is E[X]? The probability that a vacuum cleaner saleswoman makes no sales today is 1/3, the probability she makes 1 sale is 1/2, and the probability she makes 2 sales is 1/6. Each sale made is independent and equally likely to be either a standard cleaner, which costs $500, or a deluxe cleaner, which costs $1000. Let X denote the total dollar value of all sales. (a) Find P{X = 0}. (b) Find P{X = 500}. (c) Find P{X = 1000}. (d) Find P{X = 1500}. (e) Find P{X = 2000}. (f) Find E[X].

257

258

CH A P T E R 5: Discrete Random Variables

(g) Suppose that the saleswoman receives a 20 percent commission on the sales that she makes. Let Y denote the amount of money she earns. Find E[Y ]. 12. The 5 families living on a certain block have a total of 12 children. One of the families has 4 children, one has 3, two have 2, and one has 1. Let X denote the number of children in a randomly selected family, and let Y denote the number of children in the family of a randomly selected child. That is, X refers to an experiment in which each of the ﬁve families is equally likely to be selected, whereas Y refers to one in which each of the 12 children is equally likely to be selected. (a) Which do you think has the larger expected value, X or Y ? (b) Calculate E[X] and E[Y ]. 13. A ﬁnancier is evaluating two investment possibilities. Investment A will result in $200,000 proﬁt $100,000 proﬁt $150,000 loss

with probability 1/4 with probability 1/4 with probability 1/2

Investment B will result in $300,000 $200,000 $150,000 $400,000

proﬁt proﬁt loss loss

with probability with probability with probability with probability

1/8 1/4 3/8 1/4

(a) What is the expected proﬁt of investment A? (b) What is the expected proﬁt of investment B? (c) What is the investor’s expected proﬁt if she or he invests in both A and B? 14. If Var (X) = 4, ﬁnd (a) Var (2X + 14) (b) SD (2X) (c) SD (2X + 14) 15. Suppose E[X] = μ and SD(X) = σ . Let Y=

X−μ σ

(a) Show that E[Y ] = 0. (b) Show that Var (Y ) = 1. The random variable Y is called the standardized version of X. That is, given a random variable, if we subtract its expected value and divide the result by its standard deviation, then the resulting random variable

Review Problems

16.

17.

18.

19.

20.

is said to be standardized. The standardized variable has expected value 0 and variance 1. A manager has two clients. The gross annual earnings of his ﬁrst client are a random variable with expected value $200,000 and standard deviation $60,000. The gross annual earnings of his second client are a random variable with expected value $140,000 and standard deviation $50,000. If the manager’s fee is 15 percent of his ﬁrst client’s gross earnings and 20 percent of his second client’s gross earnings, ﬁnd the (a) Expected value of the manager’s fee (b) Standard deviation of the manager’s total fee In part (b) assume that the earnings of the two clients are independent. A weighted coin that comes up heads with probability 0.6 is ﬂipped n times. Find the probability that the total number of heads in these ﬂips exceeds the total number of tails when (a) n = 1 (b) n = 3 (c) n = 5 (d) n = 7 (e) n = 9 (f) n = 19 Each customer who enters a television store will buy a normal-size television with probability 0.3, buy an extra-large television with probability 0.1, or not buy any television with probability 0.6. Find the probability that the next 5 customers (a) Purchase a total of 3 normal-size sets (b) Do not purchase any extra-large sets (c) Purchase a total of 2 sets A saleswoman has a 60 percent chance of making a sale each time she visits a computer store. She visits 3 stores each month. Assume that the outcomes of successive visits are independent. (a) What is the probability she makes no sales next month? (b) What is the probability she makes 2 sales next month? (c) What is the probability that she makes at least 1 sale in each of the next 3 months? Let X be a binomial random variable such that E[X] = 6

and

Var (X) = 2.4

Find (a) P{X > 2} (b) P{X ≤ 9} (c) P{X = 12} 21. A coin that comes up heads with probability 1/3 is to be ﬂipped 3 times. Which is more likely: that heads appears exactly once, or that it does not appear exactly once?

259

This page intentionally left blank

CHAPTER 6

Normal Random Variables Among other peculiarities of the 19th century is this one, that by initiating the systematic collection of statistics it has made the quantitative study of social forces possible. Alfred North Whitehead

CONTENTS 6.1

Introduction ......................................................................... 262

6.2

Continuous Random Variables ............................................... 262 Problems ............................................................................. 264

6.3

Normal Random Variables ..................................................... 266 Problems ............................................................................. 269

6.4

Probabilities Associated with a Standard Normal Random Variable .............................................................................. 271 Problems ............................................................................. 276

6.5

Finding Normal Probabilities: Conversion to the Standard Normal ................................................................................ 277

6.6

Additive Property of Normal Random Variables ........................ 279 Problems ............................................................................. 281

6.7

Percentiles of Normal Random Variables ................................. 284 Problems ............................................................................. 289

Key Terms .................................................................................. 290 Summary .................................................................................... 290 Review Problems ......................................................................... 293

Introductory Statistics, DOI: 10.1016/B978-0-12-374388-6.00006-5 © 2010, Elsevier Inc. All rights reserved.

261

262

CH A P T E R 6: Normal Random Variables

We introduce continuous random variables, which are random variables that can take on any value in an interval. We show how their probabilities are determined from an associated curve known as a probability density function. A special type of continuous random variable, known as a normal random variable, is studied. The standard normal random variable is introduced, and a table is presented that enables us to compute the probabilities of that variable. We show how any normal random variable can be transformed to a standard one, enabling us to determine its probabilities. We present the additive property of normal random variables. The percentiles of normal random variables are studied.

6.1 INTRODUCTION In this chapter we introduce and study the normal distribution. Both from a theoretical and from an applied point of view, this distribution is unquestionably the most important in all statistics. The normal distribution is one of a class of distributions that are called continuous. Continuous distributions are introduced in Sec. 6.2. In Sec. 6.3 we deﬁne what is meant by a normal distribution and present an approximation rule concerning its probabilities. In Sec. 6.4, we consider the standard normal distribution, which is a normal distribution having mean 0 and variance 1, and we show how to determine its probabilities by use of a table. In Sec. 6.5 we show how any normal random variable can be transformed to a standard normal, and we use this transformation to determine the probabilities of that variable. The additive property of normal random variables is discussed in Sec. 6.6, and in Sec. 6.7 we consider their percentiles. The normal distribution was introduced by the French mathematician Abraham De Moivre in 1733.

6.2 CONTINUOUS RANDOM VARIABLES Whereas the possible values of a discrete random variable can be written as a sequence of isolated values, a continuous random variable is one whose set of possible values is an interval. That is, a continuous random variable is able to take on any value within some interval. For example, such variables as the time it takes to complete a scientiﬁc experiment and the weight of an individual are usually considered to be continuous random variables. Every continuous random variable X has a curve associated with it. This curve, formally known as a probability density function, can be used to obtain probabilities associated with the random variable. This is accomplished as follows. Consider any two points a and b, where a is less than b. The probability that X assumes a

6.2 Continuous Random Variables

263

FIGURE 6.1 Probability density function of X.

value that lies between a and b is equal to the area under the curve between a and b. That is, P{a ≤ X ≤ b} = area under curve between a and b Figure 6.1 presents a probability density function. Since X must assume some value, it follows that the total area under the density curve must equal 1. Also, since the area under the graph of the probability density function between points a and b is the same regardless of whether the endpoints a and b are themselves included, we see that P{a ≤ X ≤ b} = P{a < X < b}

(North Wind Picture Archives )

Historical Perspective Abraham De Moivre (1667–1754) Today there is no shortage of statistical consultants, many of whom ply their trade in the most elegant of settings. However, the ﬁrst of their breed worked, in the early years of the 18th century, out of a dark, grubby betting shop in Long Acres, London, known as Slaughter’s Coffee House. He was Abraham De Moivre, a Protestant refugee from Catholic France, and for a price he would compute the probability of gambling bets in all types of games of chance. Abraham De Moivre

Although De Moivre, the discoverer of the normal curve, made his living at the coffee shop, he was a mathematician of recognized abilities. Indeed, he was a member of the Royal Society and was reported to be an intimate of Isaac Newton.

264

CH A P T E R 6: Normal Random Variables

This is Karl Pearson imagining De Moivre at work at Slaughter’s Coffee House: I picture De Moivre working at a dirty table in the coffee house with a broken-down gambler beside him and Isaac Newton walking through the crowd to his corner to fetch out his friend. It would make a great picture for an inspired artist.

That is, the probability that a continuous random variable lies in some interval is the same whether you include the endpoints of the interval or not. The probability density curve of a random variable X is a curve that never goes below the x axis and has the property that the total area between it and the x axis is equal to 1. It determines the probabilities of X in that the area under the curve between points a and b is equal to the probability that X is between a and b.

PROBLEMS 1. Figure 6.2 is a probability density function for the random variable that represents the time (in minutes) it takes a repairer to service a television. The numbers in the regions represent the areas of those regions. What is the probability that the repairer takes (a) Less than 20 (b) Less than 40 (c) More than 50 (d) Between 40 and 70 minutes to complete a repair? 2. A random variable is said to be a uniform random variable in the interval (a, b) if its set of possible values is this interval and if its density curve is a horizontal line. That is, its density curve is as given in Fig. 6.3.

FIGURE 6.2 Probability density function of X.

6.2 Continuous Random Variables

FIGURE 6.3 Density curve of the uniform (a, b) random variable.

3.

4.

5.

6.

(a) Explain why the height of the density curve is 1/(b − a). (Hint: Remember that the total area under the density curve must equal 1, and recall the formula for the area of a rectangle.) (b) What is P{X ≤ (a + b)/2}? Suppose that X is a uniform random variable over the interval (0, 1). That is, a = 0 and b = 1 for the random variable in Prob. 2. Find (a) P{X > 1/3} (b) P{X ≤ 0.7} (c) P{0.3 < X ≤ 0.9} (d) P{0.2 ≤ X < 0.8} You are to meet a friend at 2 p.m. However, while you are always exactly on time, your friend is always late and indeed will arrive at the meeting place at a time uniformly distributed between 2 and 3 p.m. Find the probability that you will have to wait (a) At least 30 minutes (b) Less than 15 minutes (c) Between 10 and 35 minutes (d) Less than 45 minutes Suppose in Prob. 4 that your friend will arrive at a time that is uniformly distributed between 1:30 and 3 p.m. Find the probability that (a) You are the ﬁrst to arrive. (b) Your friend will have to wait more than 15 minutes. (c) You will have to wait over 30 minutes. Suppose that the number of minutes of playing time of a certain college basketball player in a randomly chosen game has the following density curve.

265

266

CH A P T E R 6: Normal Random Variables

Find the probability that the player plays (a) Over 20 minutes (b) Less than 25 minutes (c) Between 15 and 35 minutes (d) More than 35 minutes 7. Let X denote the number of minutes played by the basketballer of Prob. 6. Find (a) P{20 < X < 30} (b) P{X > 50} (c) P{20 < X < 40} (d) P{15 < X < 25} 8. It is now 2 p.m., and Joan is planning on studying for her statistics test until 6 p.m., when she will have to go out to dinner. However, she knows that she will probably have interruptions and thinks that the amount of time she will actually spend studying in the next 4 hours is a random variable whose probability density curve is as follows:

(a) What is the height of the curve at the value 2? (Hint: You will have to recall the formula for the area of a triangle.) (b) What is the probability she will study more than 3 hours? (c) What is the probability she will study between 1 and 3 hours?

6.3 NORMAL RANDOM VARIABLES The most important type of random variable is the normal random variable. The probability density function of a normal random variable X is determined by two parameters: the expected value and the standard deviation of X. We designate these values as μ and σ , respectively. That is, we will let μ = E[X] and

σ = SD(X)

The normal probability density function is a bell-shaped density curve that is symmetric about the value μ. Its variability is measured by σ . The larger σ is, the

6.3 Normal Random Variables

FIGURE 6.4 Three normal probability density functions.

more variability there is in this curve. Figure 6.4 presents three different normal probability density functions. Note how the curves ﬂatten out as σ increases. Because the probability density function of a normal random variable X is symmetric about its expected value μ, it follows that X is equally likely to be on either side of μ. That is, P{X < μ} = P{X > μ} =

1 2

Not all bell-shaped symmetric density curves are normal. The normal density curves are speciﬁed by a particular formula: The height of the curve above point x on the abscissa is 1 2 2 e−(x−μ) /2σ 2πσ

√

Although we will not make direct use of this formula, it is interesting to note that it involves two of the famous constants of mathematics: π (the area of a circle of radius 1) and e (which is the base of the natural logarithms). Also note that this formula is completely speciﬁed by the mean value μ and the standard deviation σ . A normal random variable having mean value 0 and standard deviation 1 is called a standard normal random variable, and its density curve is called the standard normal curve. Figure 6.5 presents the standard normal curve. In this text we will use (and reserve) the letter Z to represent a standard normal random variable. In Sec. 6.5 we will show how to determine probabilities concerning an arbitrary normal random variable by relating them to probabilities about the standard normal random variable. In doing so, we will show the following useful approximation rule for normal probabilities.

267

268

CH A P T E R 6: Normal Random Variables

FIGURE 6.5 The standard normal curve.

Approximation Rule A normal random variable with mean μ and standard deviation σ will be Between μ − σ and μ + σ with approximate probability 0.68 Between μ − 2σ and μ + 2σ with approximate probability 0.95 Between μ − 3σ and μ + 3σ with approximate probability 0.997 This approximation rule is illustrated in Fig. 6.6. It often enables us to obtain a quick feel for a data set.

■

Example 6.1 Test scores on the Scholastic Aptitude Test (SAT) verbal portion are normally distributed with a mean score of 504. If the standard deviation of a score is 84, then we can conclude that approximately 68 percent of all scores are between 504 − 84 and 504 + 84. That is, approximately 68 percent of the scores are between 420 and 588. Also, approximately 95 percent of them are between 504 − 168 = 336 and 504 + 168 = 672; and approximately 99.7 percent are between 252 and 756. ■

The approximation rule is the theoretical basis of the empirical rule of Sec. 3.6. The connection between these rules will become apparent in Chap. 8, when we show how a sample mean and sample standard deviation can be used to estimate the quantities μ and σ . By using the symmetry of the normal curve about the value μ, we can obtain other facts from the approximation rule. For instance, since the area between μ and μ + σ is the same as that between μ − σ and μ, it follows from this rule

6.3 Normal Random Variables

FIGURE 6.6 Approximate areas under a normal curve.

that a normal random variable will be between μ and μ + σ with approximate probability 0.68/2 = 0.34.

PROBLEMS 1. The systolic blood pressures of adults, in the appropriate units, are normally distributed with a mean of 128.4 and a standard deviation of 19.6. (a) Give an interval in which the blood pressures of approximately 68 percent of the population fall. (b) Give an interval in which the blood pressures of approximately 95 percent of the population fall. (c) Give an interval in which the blood pressures of approximately 99.7 percent of the population fall. 2. The heights of a certain population of males are normally distributed with mean 69 inches and standard deviation 6.5 inches. Approximate the proportion of this population whose height is less than 82 inches. Problems 3 through 16 are multiple-choice problems. Give the answer you think is closest to the true answer. Remember, Z always refers to a standard normal random variable. Draw a picture in each case to justify your answer. 3. P{−2 < Z < 2} is approximately (a) 0.68 (b) 0.95 (c) 0.975 (d) 0.50

269

270

CH A P T E R 6: Normal Random Variables

4. P{Z > −1} is approximately (a) 0.50 (b) 0.95 (c) 0.84 (d) 0.16 5. P{Z > 1} is approximately (a) 0.50 (b) 0.95 (c) 0.84 (d) 0.16 6. P{Z > 3} is approximately (a) 0.30 (b) 0.05 (c) 0 (d) 0.99 7. P{Z < 2} is approximately (a) 0.95 (b) 0.05 (c) 0.975 (d) 0.025 In Probs. 8 to 11, X is a normal random variable with expected value 15 and standard deviation 4. 8. The probability that X is between 11 and 19 is approximately (a) 0.50 (b) 0.95 (c) 0.68 (d) 0.34 9. The probability that X is less than 23 is approximately (a) 0.975 (b) 0.95 (c) 0.68 (d) 0.05 10. The probability that X is less than 11 is approximately (a) 0.34 (b) 0.05 (c) 0.16 (d) 0.50 11. The probability that X is greater than 27 is approximately (a) 0.05 (b) 0 (c) 0.01 (d) 0.32 12. Variable X is a normal random variable with standard deviation 3. If the probability that X is between 7 and 19 is 0.95, then the expected value of X is approximately (a) 16 (b) 15 (c) 14 (d) 13 13. Variable X is a normal random variable with standard deviation 3. If the probability that X is less than 16 is 0.84, then the expected value of X is approximately (a) 16 (b) 15 (c) 14 (d) 13 14. Variable X is a normal random variable with standard deviation 3. If the probability that X is greater than 16 is 0.975, then the expected value of X is approximately (a) 20 (b) 22 (c) 23 (d) 25 15. Variable X is a normal random variable with expected value 100. If the probability that X is greater than 90 is 0.84, then the standard deviation of X is approximately (a) 5 (b) 10 (c) 15 (d) 20 16. Variable X is a normal random variable with expected value 100. If the probability that X is greater than 130 is 0.025, then the standard deviation of X is approximately (a) 5 (b) 10 (c) 15 (d) 20 17. If X is normal with expected value 100 and standard deviation 2, and Y is normal with expected value 100 and standard deviation 4, is X or is Y more likely to (a) Exceed 104 (b) Exceed 96 (c) Exceed 100

6.4 Probabilities Associated with a Standard Normal Random Variable

18. If X is normal with expected value 100 and standard deviation 2, and Y is normal with expected value 105 and standard deviation 10, is X or is Y more likely to (a) Exceed 105 (b) Be less than 95 19. The scores on a particular job aptitude test are normal with expected value 400 and standard deviation 100. If a company will consider only those applicants scoring in the top 5 percent, determine whether it should consider one whose score is (a) 400 (b) 450 (c) 500 (d) 600

6.4 PROBABILITIES ASSOCIATED WITH A STANDARD NORMAL RANDOM VARIABLE Let Z be a standard normal random variable. That is, Z is a normal random variable with mean 0 and standard deviation 1. The probability that Z is between two numbers a and b is equal to the area under the standard normal curve between a and b. Areas under this curve have been computed, and tables have been prepared that enable us to ﬁnd these probabilities. One such table is Table 6.1. For each nonnegative value of x, Table 6.1 speciﬁes the probability that Z is less than x. For instance, suppose we want to determine P{Z < 1.22}. To do this, ﬁrst we must ﬁnd the entry in the table corresponding to x = 1.22. This is done by ﬁrst searching the left-hand column to ﬁnd the row labeled 1.2 and then searching the top row to ﬁnd the column labeled 0.02. The value that is in both the row labeled 1.2 and the column labeled 0.02 is the desired probability. Since this value is 0.8888, we see that P{Z < 1.22} = 0.8888 A portion of Table 6.1 illustrating the preceding is presented here: x

0.00

0.0 .. .

0.5000

0.5040

1.1 1.2 1.3

0.8413 0.8849 0.9032

0.8869

0.01

0.02

0.03

…

0.09

0.8888

We can also use Table 6.1 to determine the probability that Z is greater than x. For instance, suppose we want to determine the probability that Z is greater than 2. To accomplish this, we note that either Z is less than or equal to 2 or Z is greater than 2, and so P{Z ≤ 2} + P{Z > 2} = 1

271

272

CH A P T E R 6: Normal Random Variables

Table 6.1 Standard Normal Probabilities x

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.0 0.1 0.2 0.3 0.4

0.5000 0.5398 0.5793 0.6179 0.6554

0.5040 0.5438 0.5832 0.6217 0.6591

0.5080 0.5478 0.5871 0.6255 0.6628

0.5120 0.5517 0.5910 0.6293 0.6664

0.5160 0.5557 0.5948 0.6331 0.6700

0.5199 0.5596 0.5987 0.6368 0.6736

0.5239 0.5636 0.6026 0.6406 0.6772

0.5279 0.5675 0.6064 0.6443 0.6808

0.5319 0.5714 0.6103 0.6480 0.6844

0.5359 0.5753 0.6141 0.6517 0.6879

0.5 0.6 0.7 0.8 0.9

0.6915 0.7257 0.7580 0.7881 0.8159

0.6950 0.7291 0.7611 0.7910 0.8186

0.6985 0.7324 0.7642 0.7939 0.8212

0.7019 0.7357 0.7673 0.7967 0.8238

0.7054 0.7389 0.7704 0.7995 0.8264

0.7088 0.7422 0.7734 0.8023 0.8289

0.7123 0.7454 0.7764 0.8051 0.8315

0.7157 0.7486 0.7794 0.8078 0.8340

0.7190 0.7517 0.7823 0.8106 0.8365

0.7224 0.7549 0.7852 0.8133 0.8389

1.0 1.1 1.2 1.3 1.4

0.8413 0.8643 0.8849 0.9032 0.9192

0.8438 0.8665 0.8869 0.9049 0.9207

0.8461 0.8686 0.8888 0.9066 0.9222

0.8485 0.8708 0.8907 0.9082 0.9236

0.8508 0.8729 0.8925 0.9099 0.9251

0.8531 0.8749 0.8944 0.9115 0.9265

0.8554 0.8770 0.8962 0.9131 0.9279

0.8577 0.8790 0.8980 0.9147 0.9292

0.8599 0.8810 0.8997 0.9162 0.9306

0.8621 0.8830 0.9015 0.9177 0.9319

1.5 1.6 1.7 1.8 1.9

0.9332 0.9452 0.9554 0.9641 0.9713

0.9345 0.9463 0.9564 0.9649 0.9719

0.9357 0.9474 0.9573 0.9656 0.9726

0.9370 0.9484 0.9582 0.9664 0.9732

0.9382 0.9495 0.9591 0.9671 0.9738

0.9394 0.9505 0.9599 0.9678 0.9744

0.9406 0.9515 0.9608 0.9686 0.9750

0.9418 0.9525 0.9616 0.9693 0.9756

0.9429 0.9535 0.9625 0.9699 0.9761

0.9441 0.9545 0.9633 0.9706 0.9767

2.0 2.1 2.2 2.3 2.4

0.9772 0.9821 0.9861 0.9893 0.9918

0.9778 0.9826 0.9864 0.9896 0.9920

0.9783 0.9830 0.9868 0.9898 0.9922

0.9788 0.9834 0.9871 0.9901 0.9925

0.9793 0.9838 0.9875 0.9904 0.9927

0.9798 0.9842 0.9878 0.9906 0.9929

0.9803 0.9846 0.9881 0.9909 0.9931

0.9808 0.9850 0.9884 0.9911 0.9932

0.9812 0.9854 0.9887 0.9913 0.9934

0.9817 0.9857 0.9890 0.9916 0.9936

2.5 2.6 2.7 2.8 2.9

0.9938 0.9953 0.9965 0.9974 0.9981

0.9940 0.9955 0.9966 0.9975 0.9982

0.9941 0.9956 0.9967 0.9976 0.9982

0.9943 0.9957 0.9968 0.9977 0.9983

0.9945 0.9959 0.9969 0.9977 0.9984

0.9946 0.9960 0.9970 0.9978 0.9984

0.9948 0.9961 0.9971 0.9979 0.9985

0.9949 0.9962 0.9972 0.9979 0.9985

0.9951 0.9963 0.9973 0.9980 0.9986

0.9952 0.9964 0.9974 0.9981 0.9986

3.0 3.1 3.2 3.3 3.4

0.9987 0.9990 0.9993 0.9995 0.9997

0.9987 0.9991 0.9993 0.9995 0.9997

0.9987 0.9991 0.9994 0.9995 0.9997

0.9988 0.9991 0.9994 0.9996 0.9997

0.9988 0.9992 0.9994 0.9996 0.9997

0.9989 0.9992 0.9994 0.9996 0.9997

0.9989 0.9992 0.9994 0.9996 0.9997

0.9989 0.9992 0.9995 0.9996 0.9997

0.9990 0.9993 0.9995 0.9996 0.9997

0.9990 0.9993 0.9995 0.9997 0.9998

Data value in table is P{Z < x}.

6.4 Probabilities Associated with a Standard Normal Random Variable

or P{Z > 2} = 1 − P{Z ≤ 2} = 1 − 0.9772 = 0.0228 In other words, the probability that Z is larger than x can be obtained by subtracting from 1 the probability that Z is smaller than x. That is, for any x, P{Z > x} = 1 − P{Z ≤ x}

■

Example 6.2 Find (a) P{Z < 1.5} (b) P{Z ≥ 0.8} Solution (a) From Table 6.1, P{Z < 1.5} = 0.9332 (b) From Table 6.1, P{Z < 0.8} = 0.7881 and so P{Z ≥ 0.8} = 1 − 0.7881 = 0.2119

■

While Table 6.1 speciﬁes P{Z < x} for only nonnegative values of x, it can be used even when x is negative. Probabilities for negative x are obtained from the table by making use of the symmetry about zero of the standard normal curve. For instance, suppose we want to calculate the probability that Z is less than −2. By symmetry (see Fig. 6.7), this is the same as the probability that Z is greater than 2; and so P{Z < −2} = 1 − P{Z > 2} = 1 − P{Z < 2} = 1 − 0.9772 = 0.0028 In general, for any value of x, P{Z < −x} = P{Z > x} = 1 − P{Z < x}

273

274

CH A P T E R 6: Normal Random Variables

FIGURE 6.7 P{Z < −2} = P{Z > 2}.

FIGURE 6.8 P{a < Z < b} = P{Z < b} − P{Z < a}.

We can determine the probability that Z lies between a and b, for a < b, by determining the probability that Z is less than b and then subtracting from this the probability that Z is less than a. (See Fig. 6.8.) ■

Example 6.3 Find (a) P{1 < Z < 2} (b) P{−1.5 < Z < 2.5} Solution (a) P{1 < Z < 2} = P{Z < 2} − P{Z < 1} = 0.9772 − 0.8413 = 0.1359 (b) P{−1.5 < Z < 2.5} = P{Z < 2.5} − P{Z < −1.5} = P{Z < 2.5} − P{Z > 1.5} = 0.9938 − (1 − 0.9332) = 0.9270

6.4 Probabilities Associated with a Standard Normal Random Variable

FIGURE 6.9 P{Z > a} = P{Z < −a}.

Let a be positive and consider P{|Z| > a}, the probability that a standard normal is, in absolute value, larger than a. Since |Z| will exceed a if either Z > a or Z < −a, we see that P{|Z| > a} = P{Z > a} + P{Z < −a} = 2P{Z > a} where the last equality uses the symmetry of the standard normal density curve (Fig. 6.9). ■ ■

Example 6.4 Find P{|Z| > 1.8}. Solution P{|Z| > 1.8} = 2P{Z > 1.8} = 2(1 − 0.9641) = 0.0718

■

Another easily established result is that for any positive value of a P{−a < Z < a} = 2P{Z < a} − 1 The veriﬁcation of this result is left as an exercise. Table 6.1 is also listed as Table D.1 in App. D. In addition, Program 6-1 can be used to obtain normal probabilities. You enter the value x, and the program outputs P{Z < x}.

275

276

CH A P T E R 6: Normal Random Variables

■

Example 6.5 Determine P{Z > 0.84}. Solution We can either use Table 6.1 or run Program 6-1, which computes the probability that a standard normal random variable is less than x. Running Program 6-1, we learn that if the desired value of x is 0.84, the probability is 0.7995459. The desired probability is 1 − 0.80 = 0.20. That is, there is a 20 percent chance that a standard normal random variable will exceed 0.84. ■

PROBLEMS 1. For a standard normal random variable Z ﬁnd (a) P{Z < 2.2} (b) P{Z > 1.1} (c) P{0 < Z < 2} (d) P{−0.9 < Z < 1.2} (e) P{Z > −1.96} (f) P{Z < −0.72} (g) P{|Z| < 1.64} (h) P{|Z| > 1.20} (i) P{−2.2 < Z < 1.2} 2. Show that −Z is also a standard normal random variable. Hint: It sufﬁces to show that, for all x, P{−Z < x} = P{Z < x} 3. Find the value of the question mark: P{−3 < Z < −2} = P{2 < Z