Book Reviews - JISCMail

Book ReviewsThis section will review those books whose content and level re ect the general

editorial policy of Technometrics Publishers should send books for review to Eric RZiegel BP Naperville Complex Mail Code C-7 150 West Warrenville Road NapervilleIL 60563-8460 (ziegelerbpcom)

The opinions expressed in this section are those of the reviewers These opinionsdo not represent positions of the reviewerrsquos organization and may not re ect those ofthe editors or the sponsoring societies Listed prices re ect information provided by thepublisher and may not be current

The book purchase programs of the American Society for Quality can provide someof these books at reduced prices for members For information contact the AmericanSociety for Quality 1-800-248-1946

Modern Statistics for Engineering and QualityImprovement

John Lawson and John Erjavec Barry A Bodt 186

Multivariate Analysis of Quality An IntroductionHarald Martens and Magni Martens Charles K Bayne 186

SPC Practical Understanding of Capability byImplementing Statistical Process Control Third Edition

James C Abbott Phillip Yates 187

Statistical Process Control and Quality ImprovementFourth Edition

Gerald M Smith Lora Zimmer 188

Quality Improvement With Design of ExperimentsIvan N Vuchkov and Lidia N Boyadjieva Timothy Robinson 188

Six Sigma Simpli ed Quantum Improvement Made Easy

Jay Arthur Melvin Alexander 189

Design and Analysis in Chemical ResearchRoy L Tranter (editor) Margaret A Nemeth 190

A Primer for Sampling Solids Liquids and GasesPatricia L Smith Margaret A Nemeth 190

Elements of Sampling Theory and MethodsZ Govindarajulu Subir Ghosh 191

Introduction to Linear Regression Analysis Third EditionDouglas C Montgomery Elizabeth A Peck andG Geoffrey Vining J Brian Gray 191

Applied Regression Analysis for Business and EconomicsThird Edition

Terry E Dielman Henry W Altland 192

Eliciting and Analyzing Expert Judgment A PracticalGuide

Mary A Meyer and Jane M Booker J Charles Kerkering 193

Fitting Statistical Distributions The Generalized LambdaDistribution and Generalized Bootstrap Methods

Zaven A Karian and Edward J Dudewicz Thomas E Wehrly 194

Time-Series ForecastingChris Chat eld Craig B Borkowf 194

Practical Time SeriesGareth Janacek Pradipta Sarkar 195

Practical Time-Frequency Analysis Gabor and WaveletTransforms With an Implementation in S

Reneacute Carmona Wen-Liang Hwang andBruno Torreacutesani Maliha S Nash 196

Geometric Data Analysis An Empirical Approach toDiminsionality Reduction and the Study of Patterns

Michael Kirby Yachen Lin 196

Introduction to Graphical Modelling Second EditionDavid Edwards Graham J Wills 197

Editor Reports on New Editions ProceedingsCollections and Other Books

Engineering Statistics Second EditionDouglas C Montgomery George C Runger andNorma F Hubele 197

Fault Detection and Diagnosis in Industrial SystemsL Chiang E Russell and R Braatz 197

Scan StatisticsJoseph Glaz Joseph Naus and Sylvan Wallenstein 198

Advanced Linear Modeling Second EditionRonald Christensen 198

The Six Sigma RevolutionGeorge Eckes 199

Soft ComputingL Fortuna G Rizzotto M Lavorgna G Nunnari M G Xibilia and R Caponetto 199

Making Hard Decisions Second EditionRobert T Clemen and Terence Reilly 199

Applying Statistics in the CourtroomPhillip I Good 200

Statistics for Lawyers Second Edition

Michael O Finkelstein and Bruce Levin 200

Statistics for Environmental Science and ManagementBryan F J Manly 201

Bayesian Survival AnalysisJoseph G Ibrahim Ming-Hui Chen and Debajyoti Sinha 201

copy 2002 American Statistical Association andthe American Society for Quality

TECHNOMETRICS MAY 2002 VOL 44 NO 2

185

186 BOOK REVIEWS

Analyzing Medical Data Using S-PLUSB S Everitt and Sophia Rabe-Hesketh 201

Biostatistics in Clinical TrialsCarol Redmond and Theodore Colton (editors) 202

Encyclopedia of Epidemiologic Methods

Mitchell H Gail and Jacques Benichou 202

Forthcoming Reviews 203

Modern Statistics for Engineering and Quality Improve-ment by John Lawson and John Erjavec Paci c GroveCA Duxbury 2001 ISBN 0-534-19050-2 xi C 810 pp$7195

The authors set out to produce an undergraduat e text and reference thatfocuses on what they feel are the most useful statistical tools encounteredby practicing engineers They succeeded in creating a valuable book rich inexperimental design and response surface methods (458 pages 10 chapters)augmented by standard coverage of quality control (134 pages) but with onlyminimal treatment of basic probability and statistics (132 pages) This is farmore a book on experimental design application in industry than it is a gen-eral presentation of modern statistical methods for engineers as found in forexample the books by Montgomery and Runger (1999) and Ostle TurnerHicks and McElrath (1996)

The bookrsquos 5 parts and 18 chapters are as follows

Part I Introduction1 The Scienti c Method and Statistics2 Concepts of Quality Control

Part II Basic Tools3 Theoretical Background Probability and Statistics4 Descriptive Tools5 Probability Plots6 Inferential Statistics Prediction and Statistical Decision Rules

Part III Good Experiments Make for Good Statistics7 Strategies for Experimentation with Multiple Factors8 Basic Two-Level Factorial Experiments9 Additional Tools for Design and Analysis of Two-Level Factorials

10 Regression Analysis11 Multiple Level Factorial Experiments12 Screening Designs

Part IV Optimization Experiments13 Response Surface Methodology14 Response Surface Model Fitting15 Mixture Experiments

Part V Variability and Quality16 Characterizing Variability in Data17 Shewhart Control Charts18 Off-Line Quality Control and Robust Design

The book begins with an engaging introduction supported by several exam-ples that should alert engineering students that valuable information is con-tained within The book immediately challenges students to participate inengineering thought using Pareto process ow and cause-and-effec t diagramsto address defects in circuit boards Quality problems in transmission castingsand the copperplating of ceramic substrates further develop the need for under-standing the physical situation through reasoned data collection and analysisAn erroneous gure (Fig 27) and table reference (Table 23) partially con-fuses the message of the rst motivating example in Chapter 2 involvingcontrol charting and tool wear

Part II expects much of students Discrete random variables and distribu-tions appear only ve pages into Chapter 3 with the binomial distributionjust three pages later In all only 36 pages are devoted to an introduction toprobability random variables and their distributions (including the binomial

Poisson uniform exponential normal and lognormal) and the law of largenumbers and central limit theorem This streamlined presentation requires anoccasional leap for the student For example the expected value of a linearcombination of independent variables Y D aY1 C bY2 is af rmed beginningwith

E4Y 5 D E4aY1 C bY25 DZZ

4ay1 Cby25f14y15f24y25 dy1 dy20

But the only previous mention of joint probability let alone the relationshipbetween joint and marginal densities under independence comes from themultiplication rule P4AampB5 D P4A5P4B5 An equally brief but less intimi-dating presentation of summary statistics and usual graphical characterizationsof data follows in Chapter 4 Chapter 5 presents a very readable and generaltreatment of probability plots and their use and Chapter 6 gives a clear butconcise introduction

Parts III and IV are the real strength of this book supported by ampletables of design schemes in the appendixes Topics ow naturally in answer-ing the question of what needs to be done or considered next to pursue thequality engineering problem at hand The many examples are splendid Thechapter titles specify most topics but to suggest the thoroughnes s of cover-age speci c topics include PlacketndashBurman designs mixed-level fractionalfactorial designs sequential experimentation matrix algebra for regressionBoxndashCox transformations split-plot designs central composite designs andBoxndashBehnken designs

Part V concludes with more experimental designs involving variance com-ponents measurement error and staggered nested designs in Chapter 16 TwoChapters 17 and 18 continue the discussion of control charts begun in PartI and extend it to off-line quality control and robust design Chapter 18 alsoincludes parameter design experiments

In a one-semester format this book would be appropriate for upper-levelengineering students who have had some previous exposure to basic statis-tics The authors concede that Part II is streamlined to provide more timefor exploring topics of greater interest but they assert that greater detail inthat section is not required to successfully move to the latter sections of thebook Perhaps so but the instructor should be prepared to augment if neededThe authors suggest course outlines of selected chapters dependent on theengineering interest but I think a book as comprehensive as this would bebest served over two semesters As a reference for practicing engineers itworks well

Barry A Bodt

US Army

REFERENCES

Montgomery D and Runger G (1999) Applied Statistics and Probabilityfor Engineers (2nd ed) New York Wiley

Ostle B Turner K Hicks C and McElrath G (1996) Engineering Statis-tics The Industrial Experience Belmont CA Wadsworth

Multivariate Analysis of Quality An Introduction byHarald Martens and Magni Martens West Sussex UKWiley 2001 ISBN 0-471-97428-5 xx C 445 pp $135

The authors state that this book has been written to meet a need for anentry-level textbook on multivariate data analysis for the practical researcherin academia or industry The text is presented as a narrative analysis of mul-tivariate data using soft modeling The authors have little use for classicalstatistical analysis stating that (1) ldquothe book makes little or no use of tra-ditional statistical distribution theory and takes a dim view on hypothesistestingrdquo (2) ldquotraditional hypothesis testing methods in mathematical statisticsfocus more on the noise than on the interesting signal in datardquo and (3) ldquocom-puterised statistical methods for re-sampling have already reduced the needfor teaching complicated mathematical statistics to unwilling and terri ednon-statisticiansrdquo

So how do these chemometricians predict a response matrix Y from amatrix of x-predictor variables X First matrix X is represented by the prod-uct of a score matrix TA (ie ldquoArdquo principal components ) and a loading matrixP plus an error matrix Second matrix Y is modeled by the product of theTA score matrix and a loading matrix Q plus an error matrix The Y matrix isalso represented by the linear regression of X with coef cient matrix B plus


BOOK REVIEWS 187

an error matrix The authors use cross-validation to determine the optimalnumber of principal components to estimate the predictive ability of the nalmodel to assess the reliability of modeling various x- and y-variables and todetect outliers The coef cients are displayed in terms of a reliability rangeb 2s(b) where s(b) is called the estimated standard uncertainty of regres-sion coef cients determined again from cross-validation Rather than usingstatistical testing procedures most conclusions about the optimal number (ieldquoArdquo usually two) of principal components goodness of t and importantcoef cients are made by viewing plots About 60 of the book illustratesthese points by using two examples one example on viscosity measurementsand another on a mixture experiment to optimize the recipe for making cocoaOf course there are some variations on this summary but this is the authorsrsquobasic plan for analyzing almost all types of data

Some of the statistical thinking used to develop models is not followed Forexample the authorsrsquo advise that ldquothe choice of what is X and what is Y doesnot have to follow unspoken tradition that X D cause and Y D effect butthe choice is usually not very critical If in doubt the user should try to swapX and Yrdquo The authors also consider the correlation coef cient a unit-freemeasure of the goodness of t rather than a measure of linearity The conceptof degrees of freedom is considered too complex for this book Models thatinclude squared x-variables are called nonlinear models rather than second-order models and cross-products representing interaction are given minimalconsideration In the cocoa mixture experiment these chemometricians ignorethat mixture components add to 100 and blindly use a full linear model torepresent the response

Even though this is a heuristic approach to data analysis I was surprisedthat the authors did not discuss some important topics In many chemistryexperiments the number of x-variables is larger than the number of samplesin the X matrix of dimension n k with k gt n The authors mention in theappendices that the (XrsquoX) matrices can not be inverted for regression analysisand that ldquothe inversion is made in a sequence of simple stepsrdquo But there is nodiscussion in the main text about this situation Cross-validation is the maintool for evaluation but there is no discussion about the number of experi-mental runs to leave out relative to the total number of experimental runsMost important there is no discussion about software to do the recommendedsoft modeling This is particularly important because this book provides aminimal amount of mathematics to analyze the data I am not sure how anystudent will be able to solve any problems using this book The authors donot even provide any general guidelines or principles for interpreting the plotsused to evaluate all of their examples

As I read this book I wondered who was going to teach this material Icould not envision any statistician using this book as a class text and mostapplied scientists that I know prefer a more scienti c foundation for datainterpretation In some special cases the authors do allow for statisticians toanalyze data ldquoWhen working with extremely expensive dangerous or ethi-cally dif cult issues we recommend that you let a professional statisticiantake over both the planning and the data analysisrdquo

Charles K Bayne

Oak Ridge National Laboratory

The submitted manuscript has been authored by a contractor of the US Gov-ernment under Contract No DE-AC-05-00OR22725 Accordingly the USGovernment retains a non-exclusive royalty-free license to publish or repro-duce the published form of this contribution or allow other to do so for USGovernment purposes

SPC Practical Understanding of Capability by Imple-menting Statistical Process Control (3rd ed) by JamesC Abbott Easley SC Robert Houston Smith Publishers1999 ISBN 1-887355-03-0 iv C 406 pp $4995

This book now in its third edition is the nal segment of a three-part seriesdevoted to training future managers The authorrsquos WalkaboutTM series con-sists of Organize for the Ease of Doing Business Optimize your OperationStories Tools and Lessons for Using the Principles of Process Managementto Improve Your Quality and now this book According to the book jacketMr Abbott is the president of a consulting and corporate education rm Isuspect that the WalkaboutTM series is an outgrowth of his consulting expe-riences in industry His basic premise is straightforward and re ects themescommon to quality practitioners The ldquosummit of improvement mountainrdquo is

reached through process correctness consistency and capability Consistencyand capability are the bookrsquos focus areas Within the SPC framework Abbottmaintains that the control chart is the primary tactical tool for achieving con-sistency and the capability study is the primary strategic tool A shop ooroperator ought to be the primary user of a tactical tool whereas a managerwill typically use the strategic tool

The book has 10 chapters The rst two chapters focus on basic statisticsand ldquomath tricks to make SPC workrdquo The trick outlined is nothing more thanthe central limit theorem The next three chapters cover proper subgroupingand the groundwork for SPC Part of the necessary groundwork for effectiveSPC is a diagram that Abbott calls a WalkaboutTM dependency diagram whichappears to be a simple variant of a process ow diagram Three chapters aredevoted to basic control charts and how to interpret them XbarR XbarSand individualmoving range charts are detailed NP P C and U charts arediscussed in a chapter devoted to attribute charts The book concludes with achapter on capability studies and a section on integrating tactical and strategictools The capability chapter is little more than an explanation of how tocompute percent defective assuming a normal distribution with Z scores andhow to calculate a Cpk

Abbott takes a conversationa l tone throughout the book to better serve theintended audience The examples are very clear and easy to comprehend andthe graphics are large and easy to read I was also encouraged that Abbottstresses the need for tracking the ldquocentral tendencyrdquo (mean) variability anddistribution (shape) of both process and product variables He is very detailedin presenting the needed calculations I found the level of detail slightlyannoying given the simplicity of the calculations and the availability of mod-ern software nonetheless personal experience has taught me that calculatingmeans and standard deviations is not something to be taken for granted A nicefeature used throughout the book is the insertion of ldquoQuality Hellrdquo paragraphs These are practices that we in industry routinely use to our own detriment

Despite my previous comments I cannot recommend this book Abbott haschosen to write a book about SPC when he does not appear to have a suf cientcommand of the technical details associated with basic SPC and statisticalmethods In an example for calculating probabilities associated with the ageof previous US presidents he makes the following statement ldquoA person of61 years of age and older has a probability of 1587 chance out of 100 tobecome president based on past historyrdquo (p 54) Abbott also claims a kind ofdistribution transitive property as evidenced by the following ldquoThe Poissondistribution is a special kind of binomial distribution and the binomial is aspecial kind of normal distributionrdquo (p 283) This transitivity provided thejusti cation for the attribute control charts Abbott states that con dence limitsare the basis for a tactical toolrsquos limits Similarly average plusminus threestandard deviation prediction limits are the strategic tool limits for assessingcapability ldquoThe control limits formula uses con dence intervalsrdquo (p 162)where the standard error multiplier ldquomust be set only at a Z value of THREErdquo(p 116) Abbottrsquos attempt to distinguish between the distribution of a randomvariable and its sample average has good intentions but misses the mark forcorrectness A distribution check is accomplished by a visual inspection ofa histogram To determine when a control chart alarms Abbott includes hisversion of Western Electric rules that include a series of points outside thecontrol limits more than 68 of the points lying within one standard errorof the mean and more than 32 of the points lying outside one standarderror of the mean (p 206) How to implement the last two rules on the shop oor would be a challenge without considering whether or not they makesense Walter Shewhartrsquos last name appeared three times in the book eachtime spelled ldquoSkewhartrdquo (eg p 156) I also found the book to be veryrepetitive Four of the more popular clip art graphics were used a total of21 times Abbottrsquos 1st Principle of Process Management is explicitly statedthree times and the 2nd Principle of Process Management at least four timesI would venture to guess that a discriminating editor could trim the bookto at least a quarter of its current length and still retain the authorrsquos intentExperience has taught me that both managers and those who see themselvesas future managers like books that are short and to the point Abbott alsosupplies a glossary that contains several blatant technical mistakes Examplesinclude analysis of variance the comparative study of discrete group databinomial having two modes or centers degrees of freedom the small samplecorrection formula for standard deviation (nƒ1) Rbar the central tendency ofthe variability of a metric variance (s2) the range average and Xdoublebar designed to detect movement or change in the central tendency of the processThe author provides an index at the end of the book that I suspect most would nd annoying It appears that he used the nd feature common to modern


188 BOOK REVIEWS

word processing software to document the occurrence of selected key wordsFor example the index contains 125 page references for average 216 pagereferences for control and 153 page references for control chart

I do not think the average Technometrics reader will nd this book veryuseful I could even argue that the bookrsquos intended audience of future man-agers would reap only marginal bene ts from reading this book The book iselementary repetitious and sprinkled with occasional disturbing inaccuraciesThe authorrsquos claim that this book ldquois the most rigorous and thorough book onthe topic of SPCrdquo (p 7) is not just an ambitious overstatement but wrong

Phillip Yates

In neon TechnologiesmdashRichmond

Statistical Process Control and Quality Improvement(4th ed) by Gerald M Smith Upper Saddle River NJPrentice-Hall 2001 ISBN 0-13-025563-7 xv C 650 pp$7333

This book fully covers the implementation and use of statistical processcontrol (SPC) It begins by discussing managerial aspects and then leads intothe detailed calculations required for using control charts The author statesin the rst paragraph that this latest edition was prepared with comments andsuggestions from users of previous editions and then lists some of the speci cchanges The target audience is 2- and or 4-year college students and industrialpractitioners To make this book easy to use for all of these groups the bookis mathematically friendly and uses only basic mathematics In the Prefacethe author provides a sequence for using the book based on the target audi-ence Professors instructors and trainers would nd this feature helpful Eachchapter includes objectives examples and exercises Many chapters includeat least one case study Also provided are blank control master forms thatmay be used for homework or projects The book does not include any soft-ware output or make recommendation s about the use of software thereforeall examples are calculated using a calculator and the control chart mastersprovided

Statistical Process Control and Quality Improvement is very comprehensivein its coverage of the implementation of SPC The rst chapter ldquo Introduc-tion to Quality Concepts and Statistical Process Controlrdquo gives de nitionsof quality and discusses the difference between prevention and detection thegoals of SPC basic tools for SPC and designed experiments and how theycan be used to implement SPC into an existing process The second chapterldquoStriving for Quality Managementrsquos Problem and Managementrsquos Solutionrdquodiscusses managementrsquos problem with why SPC does not always work easilythe rst time management rsquos dilemma leadership by management Demingrsquoscontribution to quality (including his 14 points for management) Crosbyrsquosapproach to quality improvement (including his 14 steps) a comparison ofthe two approaches total quality management (TQM) the Malcolm BaldridgeNational Quality Award total customer satisfaction ISO-9000 and the servicesector Three case studies are included in this chapter

Chapter 3 ldquo Introduction to Variation and Statisticsrdquo provides the statisticalbasics needed to use and understand control charts This chapter includesde nitions of accuracy maximum error tolerance distribution special-causeand common-cause variation (with a case study) variation concepts (locationspread and shape) population sample mean median mode standard devi-ation and variance The chapter also includes many gures and examples toillustrate these concepts Chapter 4 ldquoOrganization of Data Introduction toTables Charts and Graphsrdquo gives step-by-step examples for creating stemplots (often called stem-leaf plots) frequency distributions and tally chartshistograms (with a case study) Pareto charts owcharts storyboards cause-and-effect diagrams checksheets and scatterplots As stated earlier thesecharts and graphs were created manually because no software is used in thebook Chapter 5 ldquoThe Normal Probability Distributionrdquo provides basic infor-mation about probability distributions the normal distribution and applicationof the central limit theorem Numerous examples and gures clearly illustratethese concepts

Chapter 6 ldquoIntroduction to Control Chartsrdquo begins the step-by step instruc-tions for implementing and using many different types of control charts Thecontrol-charting concept is de ned and preparation for control charting dis-cussed followed by detailed instructions (11 steps) on how to implement Nxand R control charts Also included in this chapter are guidelines on whento recalculate control limits capability analysis (Cr Cp and Cpk) a de ni-tion of six-sigma quality and a case study Chapter 7 ldquoAdditional Control

Charts for Variablesrdquo discusses median and range ( Qx and R) control chartsNx and s charts coding data a modi ed Nx and R chart for small datasets thenominal Nx and R chart the transformation Nx and R chart and control chartselection Chapter 8 ldquoVariables Charts for Limited Datardquo discusses precon-trol or rainbow control charts compound probability modi ed precontrol fortight measurements and charts for individual measurements with a case studyChapters 7 and 8 include step-by-step instructions and many examples to helpthe user

Chapter 9 ldquoAttributes Control Chartsrdquo introduces the four types of attributecontrol charts (p np c and u charts) Each of theses control chart typesis presented clearly with examples In this chapter the author mentions theuse of computers for making and using control charts He also points out thedrawbacks including that they may be ldquointimidating and confusingrdquo Chap-ter 10 ldquoInterpreting Control Chartsrdquo teaches the user to distinguish betweenrandom patterns and patterns that indicate that there is a problem This chap-ter includes lots of gures and examples to show how to recognize patternshow to use probability to recognize a problem and how to use the tools inproblem solving Chapter 11 ldquoProblem Solvingrdquo further details the steps usedfor effective and ef cient problem solving These include the sequence team-work brainstorming tools mistakeproo ng problem solving in management(with a case study) just-in-time (J IT) and problem solving in the classroom(with a case study)

Chapter 12 ldquoGauge Capabilityrdquo provides the details for a gauge capabil-ity study Included are preparations a 15-step procedure with a worksheetanalysis of repeatability and reproducibility with accuracy and stability andthe elimination of gauge variation Chapter 13 ldquoAcceptance Samplingrdquo thebookrsquos nal chapter discusses methods used to sample the process It includesrandom sampling operating characteristic curves the average outgoing qual-ity curve MIL-STD-105D for inspection by attributes average proportiondefective and vendor certi cation There are four appendixes ldquoBasic MathConcepts and Probabilityrdquo ldquoCharts and Tablesrdquo ldquoGlossary of Symbolsrdquo andldquoLab Exercises for Each Chapterrdquo

This book is reader friendly and has many step-by-step examples that willbe helpful to students and practitioners using this book to learn or implementSPC It is (as stated in the Preface) also mathematically friendly and all of themathematical or statistical concepts are clearly de ned and explained Giventhe wide range of information on the implementation and use of SPC and theclear way that it is presented I believe that this book would be useful as atextbook for a college quality control course internal company training (atany level) or as a reference for practitioners implementing SPC Althoughthe author presents his reasoning for preferring the use of manually generatedcontrol charts I feel that the book would only be enhanced by the additionof some information about software currently being used in industry for SPCIn short I would recommend this comprehensive book for use in teaching orimplementing SPC

Lora Zimmer

Arizona State University

Quality Improvement With Design of Experiments byIvan N Vuchkov and Lidia N Boyadjieva Dordrecht The Netherlands Kluwer Academic Publishers 2001ISBN 0-7923-6827-4 xvi C 505 pp $190

This book provides a comprehensive treatment of robust parameter designRobust parameter design an off-line quality control method emphasizes theproper choice of levels of controllable factors (parameters) in a manufacturingprocess The choice of levels depends to a large extent on the variabilityaround some prechosen target for the production process Robust parameterdesign has received much attention from quality engineers and statisticianssince the work of Genichi Taguchi in the 1980s (see Taguchi and Wu 1980Taguchi 1986 1987) Here the authors adopt a model-building approach torobust parameter design and motivate the need for a model-based approach

The book comprises 10 chapters Chapter 1 is introductory the authorsdiscuss some fundamental terminology and philosophy on which the book isbased They mention that there are various approaches to robust parameterdesign but that their emphasis is on the Taguchi method and a model-basedapproach using response surface methods (RSMs)

In an effort to make the book self-contained Vuchkov and Boyadjievadevote close to 40 of the text to background material needed to understand


BOOK REVIEWS 189

RSMs This material is found in Chapters 2 and 3 Chapter 2 begins with anice overview of ANOVA and the de nition of factorial designs The authorsthen move into a discussion of various strategies for collecting data suchas the use of completely randomized designs randomized complete blockdesigns Latin squares Graeco-Latin squares and incomplete block designsThey then provide a solid background of regression analysis within the contextof RSMs The discussion is within a matrix framework Chapter 3 is devotedto the actual design of experiments The authors take the reader through thenotion of sequential experimentation and concepts such as screening experi-ments design resolution steepest ascent second-order designs design opti-mality and other fundamental concepts of RSMs Concepts are discussedtersely and a reader unfamiliar with the material would nd it dif cult toextract a good understanding of the topic based on this text alone Very fewexamples are presented and the discussion is often more complicated thanneeded I was disappointed that the authors did not give much discussion ofthe importance of resolution III designs

After spending considerable time on background material the authors pro-vide an overview of Taguchirsquos approach to quality improvement in Chapter 4They discuss Taguchirsquos loss function crossed arrays signal-to-noise ratiosanalysis of data and the nal decision making process The chapter ends withtwo examples I have two criticisms of this chapter First the authors fail topoint out that the problem with crossed arrays is that too many degrees offreedom are used up for controlnoise interactions and not enough are left forcontrolcontrol and noisenoise interactions Second the authors introducetwo variance components in the discussion of the analysis and their treat-ment of these components is confusing Nonetheless the authors should becommended for their discussion on the use of split-plot designs when thereare factors whose levels are hard to control (Sec 46)

In Chapter 5 the authors consider the situation in which the performancevariablersquos (response variable) variability is due to errors in factors The authorscontend that although we can often conduct an experiment without errors inthe factors it is practically impossible to organize a production process with-out such errors The authors refer to the errors-in-variables model (see Myersand Montgomery 1995) as the mass production model Mean and variancemodels are explicitly given for the two-factor case and are extended to morethan two factors in a matrix development A sound treatment of the estimationof the moments of the errors in the factors is also provided Chapter 5 endswith a detailed presentation of the potential inaccuracy of the prediction modelwhen using the mass production model A handful of examples are scatteredthroughout to illustrate the concepts

Chapter 6 is devoted to optimization techniques for the types of modelsdiscussed in Chapter 5 Optimality criteria are developed and optimizationprocedures for the ldquotarget is bestrdquo and ldquolargersmaller is betterrdquo cases arediscussed The authors also spend some time on situations involving morethan one performance characteristic

In Chapter 7 the authors consider cases involving both noise variables anderrors in design factors The concepts are well developed from a mathematicalperspective and the examples given help clarify the discussion The chapterends with optimization procedures for the considered scenario The authorsdevelop the mean and variance models in a mathematically rich appendix abetter approach would have been to develop these models within the contextof the chapter

Vuchkov and Boyadjieva address quality improvement through mechanisticmodels in Chapter 8 Examples illustrating the various topics in this chapterare well chosen but could use more elaboration In Chapter 9 the authorsconsider models for quality improvement of products andor processes thatdepend on both quantitative and qualitative factors In nal chapter 10 theycover model building (mean and variance) when there are replicated observa-tions at the design points They also describe methods of determining locationand dispersion effects from nonreplicated observations

This book is interesting and provides a nice resource for understandingmany of the issues confronting robust parameter design The authors mentionthat this work stems from industrial short courses that have been taught inthis area but the writing style is more technical rather than geared towardthe practitioner The examples used are generally well chosen although itwould have been nice had the examples been more tightly woven within themethodologica l discussion The intended audience is engineers and statisti-cians working in the eld of quality improvement If I would have had theopportunity to provide input before publication I would have suggested that

the authors spend some time mentioning available software for the method-ologies discussed Overall this book lls a valuable niche among qualityimprovement texts

Timothy Robinson

University of Wyoming

REFERENCES

Myers R H and Montgomery D C (1995) Response Surface MethodologyProcess and Product Optimization Using Designed Experiments New YorkWiley

Taguchi G (1986) Introduction to Quality Engineering White Plains NYUNIPUBAsian Productivity Organization

(1987) System of Experimental Design Engineering Methods toOptimize Quality and Minimize Cost White Plains NY UNIPUB KrausInternational

Taguchi G and Wu Y (1980) Introduction to Off-Line Quality ControlNagoya Japan Central Japan Quality Control Association

Six Sigma Simpli ed Quantum Improvement MadeEasy by Jay Arthur Denver CO LifeStar 2001 ISBN1-884180-13-2 127 pp $2495

This book begins with the statement ldquoThis QI Coloring Book is designedto make learning the principles and processes of Six Sigma more easyrdquo Assuch it sounds like a book for elementary or middle-school students ratherthan one directed toward the technical level of Technometricsrsquo readers Thisunintimidating style may be most appropriate for the ldquomathematically chal-lengedrdquo looking for some understanding of Six Sigma The author claims thatreaders will discover the essence of Six Sigma and how to implement SixSigma to maximize the gain minimize the pain and focus on creating resultsfrom the very rst day

Unfortunately in making Six Sigma easier to understand the author may befalling back on traditional total quality management (TQM) techniques ratherthan using true Six Sigma techniques For example in his problem-solvingprocess the author uses worksheets and descriptive examples to guide readersthrough his own brand of the plan-do-check-ac t (PDCA) cycle which hecalls FISH (focus-improve-sustain-honor) FISH attempts to substitute for themore familiar de ne-measure-analyze-improve-contro l (DMAIC) Six Sigmaimprovement cycle but is actually closer to the TQM philosophy than to theDMAIC approach considered the standard road map strategy throughout muchof the Six Sigma literature

Speci cally the author makes no mention of measurement systems studies(the ldquoMrdquo phase in DMAIC) and provides no examples of completed projectsor case studies that show how his ldquoSix Sigmardquo approach has been effectiveHe could have shown the merits of his Microsoft Excel software templatesmore effectively by using them to complete examples illustrating how the SixSigma tools can be used to actually analyze and solve real problems Theabsence of any completed examples does little to give readers any con dencethat these software templates are credible or worth the money Instead thisbook is more like an advertisement for the authorrsquos software templates than ahow-to book or even a resource for references where readers can obtain moreinformation

This publication may be best suited for trainers endeavouring to acquaintemployees with quality and business management tools in a way that allowsfor a gentle transition into the more complex mathematical methods of datacollection The book provides a clear way to select and use control charts thatcan be understood by colleagues who admit to being afraid of math Howeverthis book is not truly Six Sigma simpli ed but rather is more like TQMwith some Six Sigma avor This book appears to be primarily a marketingtool for promoting the authorrsquos particular quality improvement software yetit does not appear to do anything more beyond what other books and existingsoftware packages already do (see Taft and Amazoncom 2001) This bookcould have been more useful if it were pocket-sized like The Memory Joggerseries (GOALQPC 1999) or The Six Sigma Pocket Guide (Rath and Strong2000) In fact more technically inclined users may nd The Six Sigma PocketGuide a better buy ($1200 vs $2495) because it serves as a handy referenceguide for Six Sigma and more accurately describes the DMAIC tools and howthey can be applied

Melvin Alexander

Qualistics


190 BOOK REVIEWS

REFERENCES

GOALQPC (1999) The Memory Jogger Memory Jogger II Team MemoryJogger Project Management Memory Jogger The Creativity Tools MemoryJogger Methuen MA Author

Rath and Strong (2000) Six Sigma Pocket GuideTaft W and Amazoncom (2001) Reviews of Six Sigma Simpli-

ed by Jay Arthur available at httpwwwisixsigmacomforum andhttpwwwamazoncom

Design and Analysis in Chemical Research edited byRoy L Tranter Boca Raton FL CRC 2000 ISBN0-8493-9746-4 xviii C 558 pp $13995

This book contains a series of chapters written by different individualsThe editor states in the Preface that ldquothe aim of this book is to show thatit [real statistics] is essentially an extension of the logical processes used bychemists every day and that its use can and does bring greater understandingof problems more quickly and easily then the purely intuitive or lsquoletrsquos try andseersquo approachesrdquo

The bookrsquos chapters and their contents are as follows

1 Statistical Thinking by M Porter sampling variability probabilityexperimental planning and statistical signi cance

2 Essentials of Data Gathering and Data Descriptions by R Trantermeasurement systems graphics and signi cant digits

3 Sampling by J Thompson sample size estimation limit of detection(LOD) limit of quantitation (LOQ) acceptance sampling

4 Interpreting Results by M Gerson Standard errors con dence inter-vals t tests sample size computations using con dence intervals equivalencetests

5 Robust Resistant and Nonparametric Methods by J Thompson non-parametric and robust estimates outliers nonparametric regression

6 Experimental Design by S Godbert Two-level designs and theirfractions confounding irregular designs PlackettndashBurman Taguchi andD-optimal designs

7 Designs for Response Surface Modelling by J Langhans central com-posite BoxndashBehnken uniform precision rotatability robust designs opti-mization is barely mentioned

8 Analysis of Variance by M Porter Simple linear regression one- andtwo-way analysis of variance blocking factorials nested factors no discus-sion on multiple comparisons apart from individual and simultaneous errorrates reader is referred to references

9 Optimization and Control by T Kourti SPCmdashxbar and R charts uni-variate and multivariate cusum and exponentially weighted moving average(EWMA) Hotellingrsquos T2 and chi-squared multivariate charts latent variablesprincipal component analysis (PCA) and partial least squares (PLS) Cusumuses the v-mask two one-sided cusums are not discussed Also there is hardlyany discussion on individual control charts which is surprising becausechemical manufacturing uses individual x charts quite extensively

10 Grouping Data TogethermdashCluster Analysis and Pattern Recognitionby W Melssen PCA nonlinear mapping neural networks clustering classi- cation

11 Linear Regression by R Tranter simple linear regression errorsin variables multiple linear regression stepwise PCR and PLS nonlinearregression neural networks genetic algorithms

12 Latent Variable Regression Methods by O Kvalheim Multiple linearregression PCR PLS Good basic comparison of multiple linear regressionPCR and PLS

13 Data Reconstruction Methods for Data Processing by A G Ferrigeand M R Alecio discussion of methods for tting a theoretical model to theexperimental data

What I liked about this book is that it covers a vast array of statisticalmethods and techniques and each chapter has an introduction section and aldquowhere to gordquo section that delineates the different concepts and points thereader to various sections within the chapter

What I did not like about this book is that there are not enough examplesthere are only extremely brief discussions of some concepts (the reader isgiven references) there is no discussion of statistical software and the differ-ent chapters are based on statistical concepts (what I would call the statisticalview) but not on the applied chemist view I have worked with chemists for

more than 16 years and I do not believe this book will achieve the aim givenin the preface If I were a chemist with as the editor says in the Prefacethe attitude that ldquowithin the chemical sciences statistics has the reputation ofbeing hard and usable only by mathematicians or masochistsrdquo I donrsquot believethis book would change my opinion

The editor could have better achieved the bookrsquos aim by structuring eachchapter to revolve around a particular areatype of chemistry with a discussionof the pertinent statistical methods Then each chapter could have concludedwith the analysis of real datamdashfor example in process chemistry the discus-sion of response surface designs and optimization in formulation chemistrythe discussion of mixture designs in analytical chemistry the discussion ofregression modeling and variance component estimation for near infrared(NIR) methods the discussion of partial least squares and other latent variabletechniques

Margaret A Nemeth

Monsanto Company

A Primer for Sampling Solids Liquids and Gasesby Patricia L Smith Philadelphia S IAM 2001 ISBN0-89871-473-7 xix C 96 pp $45

This book is a primer based on the sampling theory (seven sampling errors)of Pierre Gy As the author states in the Preface ldquomy graduate education insampling theory addressed only situations in which the sampling units werewell-de ned (people manufactured parts etc) I did not learn in these statis-tics courses how to sample bulk materialsrdquo I also have found this to be thecase and wish I had known about Gyrsquos theory when asked several years ago todevelop a sampling plan for tank cars containing a liquid chemical Needlessto say when I saw the advertisement for this book I quickly purchased itand have found it to be extremely useful

The author also states in the Preface that ldquothis book is an introductiononlyrdquo Nonetheless I found it to be an excellent introduction not only to Gyrsquostheory but also to a logical approach to sampling bulk products Chapter 1ldquoSome Views of Samplingrdquo is an introduction to basic sampling issues andGyrsquos structured approach Chapter 2 ldquoThe Material Sampling and SamplingVariationrdquo discusses the different types of heterogeneity that can be foundin samples As an aside I found it easier to read Chapter 2 after readingAppendix A which de nes Gyrsquos seven basic sampling errors

Chapter 3 ldquoThe Tools and Techniques Sampling Sample Collection andSample Handlingrdquo discusses different ways to sample (eg one-dimensional two-dimensional and three-dimensional sampling) along with such issues ascontamination Chapter 4 ldquoThe Process Sampling and Process Variationrdquodiscusses variation over time both short-range and long-range The two maintools introduced are time plots and variograms The variogram can be used toidentify cyclic trends and to determine how much of the variability is due tothe process and how much is due to other sampling errors and the analyticalerror

Chapter 5 ldquoA Strategy Putting Theory Into Practicerdquo is essentially a log-ical plan for sampling This plan comprises the three Arsquosmdashaudit assessmentand action In the audit step not only is the written sampling plan reviewedbut also a walk-through of the sampling area is done The assessment stagebrings in experimental design if necessary

The book also contains several appendixes Appendix B gives the formulafor the variance of the smallest sampling-to-sampling variation that is thevariance of the fundamental error Appendix C discusses sequential randomsampling Appendix D shows how to calculate the variogram and gives visualbasic code for the calculations Appendix E gives three simple experimentsthat illustrate the problems encountered when sampling three- two- and one-dimensional situations and when sampling solids and liquids Appendix Fdiscusses sampling safety

In conclusion I highly recommend this book not only to statisticians butalso to anyone involved in the sampling of bulk materials

Margaret A Nemeth

Monsanto Company


BOOK REVIEWS 191

Elements of Sampling Theory and Methods byZ Govindarajulu Upper Saddle River NJ Prentice-Hall1999 ISBN 0-13-743576-2 xvi C 416 pp $76

The author states in the Preface ldquoOne cannot do justice to sampling theoryand methods in one semester however by using books like Hansen Hurwitzand Madow (1953) Sukhatme et al (1984) and Cochran (1977) which aredesigned for a two-semester course Hence I developed this book whichevolved after teaching a course to the graduate students at the Universityof Kentucky over several years It contains as many sampling ideas at anelementary level as can be imparted during a semesterrsquos course The book isself contained in the sense that all elementary proofs are providedrdquo

An interesting feature of this book is that it presents ldquovarying-probabilit ysamplingrdquo in Chapter 2 and then revisits it again in Chapter 10 in more detailThe author is to be commended for presenting a lot of material that is notavailable in a single book Of the many recent books available in surveysampling I have used in my teaching or read those of Hedayat and Sinha(1991) Thompson (1992) Thompson (1997) and Lohr (1999) I nd all ofthese to be different from each other and furthermore they are all valuable inmy understanding of this important area of statistics Elements of SamplingTheory and Methods is unique in its presentation of materials and I like itwith the four other books mentioned earlier

The author states in the Preface that ldquoowing to the limited scope of thisbook it is different to do justice to the numerous contributors to this eld Theselection of topics is bound to be subjective and is dictated by the level of thecourse Thus the bibliography is also highly selective Hence my apologiesto those whose papers are not cited in the bibliographyrdquo This is of courseunderstandable but again the author has done a splendid job The bookrsquos priceis reasonable in comparison to the other four books mentioned in this area Istrongly recommend this book to the readers of Technometrics

Subir Ghosh

University of California Riverside

REFERENCES

Hedayat A S and Sinha B K (1991) Design and Inference in FinitePopulation Sampling New York Wiley

Lohr S (1999) Sampling Design and Analysis Belmont CA DuxburyThompson M E (1997) Theory of Sample Surveys London Chapman and

HallThompson S K (1992) Sampling New York Wiley

Introduction to Linear Regression Analysis (3rd ed)by Douglas C Montgomery Elizabeth A Peck andG Geoffrey Vining New York Wiley 2001 ISBN0-471-31565-6 xvi C 641 pp $9495

In its previous two editions Introduction to Linear Regression Analysisserved for nearly two decades as a vehicle for teaching regression analysis toupper-level undergraduate s and graduate students in various areas includingbusiness engineering and the sciences Reviews of the rst edition (Cass1983 Sampson 1985) and the second edition (Assuncao and Sampson 1993Barbur 1994) are cited in the reference list at the end of this review

The bookrsquos most similar competitor is Applied Linear Regression Models(Neter Kutner Nachtsheim and Wasserman 1996) Both books are consideredldquostandardsrdquo for teaching regression analysis and are very thorough in theirclassical coverage of regression analysis The latest editions of both bookshave added chapters on more advanced topics in regression analysis Otherstrong candidates for teachinglearning regression analysis include the booksby Cook and Weisberg (1999) and Hamilton (1992) both of which featuremore graphically oriented approaches

This new edition contains welcome improvement s to the earlier versionsincluding the addition of a third author Geoff Vining to the team of Mont-gomery and Peck and an increase in the number of chapters from 10 to 15The rst nine chapters represent the core of a one-semester course in regres-sion analysis The later chapters provide a more detailed treatment of topicsdiscussed in the rst nine chapters and give thumbnai l sketches of advancedregression topics not considered in the earlier chapters

The book gets off to a good start with a ldquobig-picturerdquo overview of regres-sion analysis in Chapter 1 Chapter 2 is a traditional treatment of simple

linear regression including derivation of the ordinary least squares (OLS)parameter estimates properties of the OLS sampling distributions con denceintervals and hypothesis testing Calculations are illustrated throughout Chap-ter 2 using the data from one example This chapter also includes sectionson the derivation of the maximum likelihood estimation (MLE) parameterestimates regression through the origin and a discussion of the case wherethe predictor variable itself is random

Chapter 3 motivates the study of multiple regression with graphical exam-ples of regression models involving two predictor variables including exam-ples with power and interaction terms The matrix representation of the regres-sion model is used in OLS and MLE derivations and in establishing propertiesCoverage within the chapter includes hypothesis testing of individual coef- cients and groups of coef cients using the extra sum of squares principleThere are also optional sections on topics usually covered in a linear modelscourse including the vector space geometry of least squares and the generallinear hypothesis Joint and individual con dence intervals and predictionintervals are also covered and a section on standardized regression coef -cients is included

The analysis of residuals to detect model violations and outliers is the topicof Chapter 4 The material covers various forms of scaled residuals includ-ing standardized internally studentized externally studentized and predictionerror sum of squares (PRESS) residuals The residual plots described includenormal probability plots plots of residuals versus predictors in or out of themodel plots of residuals versus tted values time sequence plots of residu-als partial regression plots and partial residual plots There is a section onhypothesis tests based on residuals including the PRESS statistic and lack-of- t tests (both with and without replicates) Examples are given throughoutthe chapter based on data carried over from earlier chapters

Nonlinear transformations of the regression variables to stabilize the errorvariance and to linearize the model are discussed in Chapter 5 The authorscover this topic particularly well in the current and previous editions of thebook Guidelines are given for choosing the transformation and graphicallychecking the results of a transformation Analytical procedures are also givenincluding the BoxndashCox and BoxndashTidwell methods Generalized and weightedleast squares are also discussed as alternatives for handling nonconstant errorvariance

Chapter 6 covers a variety of single-case leverage and in uence mea-sures for detecting unusual cases Notably missing from this chapter are SASand Minitab output in the examples and instructions on how to obtain casediagnostics in SAS and Minitab On the positive side there is an excellentdiscussion on the treatment of in uential data (ie considerations in keep-ing or deleting the unusual data) There is also a brief section on detectingin uential groups of data

The next two chapters are on polynomial regression models and indicatorvariables The polynomial regression chapter contains a lot of material someof which might be skipped in a one-semester course The chapter on indicatorvariables is extremely well written and covers all of the bases The authorspoint out potential problems and pitfalls in the use of regression analysisthroughout including outliers in uential data hidden extrapolation and mul-ticollinearity Later chapters provide more detailed treatments of these topics

The nal chapter of the core of nine chapters on variable selection andmodel building provides a good discussion of the problems associated withmodel misspeci cation to help motivate the need for variable selection fol-lowed by a section on criteria for comparing models Computational methods(all possible subsets forward selection backward elimination and stepwiseregression) are then described and one example is given to demonstrate themethods SAS and Minitab output are shown My complaint with this chapteris that it seems too short for the topic More examples are needed (egto give examples of modeling interactions and using indicator variables) Iwould also prescribe some initial plotting of the data to develop a feel forthe relationships among the variables and to identify any unusual data beforeattempting to t any models

The remaining six chapters cover some topics presented earlier in moredetail and also introduce some advanced topics in regression analysis Chap-ter 10 on multicollinearity is much longer than the earlier chapter on variableselection and model building There is a good discussion of sources andeffects of multicollinearity and a lot of detail on multicollinearity diagnosticsincluding variance in ation factors and condition indices and possible reme-dies for the multicollinearity problem Readers should have good grasp ofmatrix algebra (in particular eigenanalysis) to derive maximum bene t fromthis chapter


192 BOOK REVIEWS

Chapter 11 on robust regression provides fairly thorough coverage ofrobust and resistant regression for an introductory book in regression analysisTopics include M-estimation least median of squares least trimmed sum ofsquares MM-estimators (GM-estimators) compound (two-stage) estimatorsR-estimators L-estimators and robust ridge estimation There are numerousreferences to more detailed descriptions in other sources for the interestedreader

Chapter 12 is a good introduction to nonlinear regression Chapter 13 coversgeneralized linear models (GLMs) This chapter starts with binary logistic andPoisson regression models then shows how they t into the GLM frameworkThe chapter provides compact coverage of basic GLM terminology and anexample using PROC GENMOD in SAS

Thumbnail sketches of autocorrelated errors measurement errors in predic-tors inverse regression (calibration) bootstrapping in regression classi cationand regression trees (CART with the emphasis on regression) neural net-works and experimental design in regression are covered in Chapter 14 Theinterested reader will want to consult the references for more detailed treat-ments of these topics

The nal chapter is a brief discussion of validating regression models Thischapter would probably t better after Chapter 9 because it is an importanttopic and should be in the core set of chapters

Introduction to Linear Regression Analysis is particularly well suited as atext for an upper-level undergraduat e or graduate-leve l course in regressionanalysis The reader should have a basic knowledge of matrix algebra (whichcould be taught along the way in a course with perhaps an introduction tomatrix algebra at the beginning) A basic knowledge of mathematical statisticswould also be helpful in deriving maximum bene t from the book

Exercises at the ends of chapters provide a mix of applied and theoreticalproblems leaning more toward the applied Appendixes include statisticaltables lists of datasets used in the book and supplemental technical materialcontaining some matrix derivations not contained in the main text Datasetsfrom the book and extensive problem solutions are available from an ftp siteThe instructorrsquos manual contains solutions to all exercises electronic versionsof all datasets and questions and problems for use on examinations Thestudent solutions guide provides complete solutions to selected problems Ref-erences to a substantial number of books and journal articles in the regressionliterature are provided at the end of the book

I highly recommend the third edition of Introduction to Linear RegressionAnalysis to practitioners students and teachers of regression analysis

J Brian Gray

The University of Alabama

REFERENCES

Assuncao R and Sampson P D (1993) Review of Introduction to LinearRegression Analysis (2nd ed) Journal of the American Statistical Associ-ation 88 383

Barbur V A (1994) Review of Introduction to Linear Regression Analysis(2nd ed) Statistician 43 339ndash341

Cass J M (1983) Review of Introduction to Linear Regression AnalysisApplied Statistics 32 94ndash95

Cook R D and Weisberg S (1999) Applied Regression Including Comput-ing and Graphics New York Wiley

Hamilton L C (1992) Regression With Graphics A Second Course inApplied Statistics Paci c Grove CA Brooks-Cole

Neter J Kutner M H Nachtsheim C J and Wasserman W (1996)Applied Linear Regression Models (3rd ed) Chicago Irwin

Sampson P D (1985) Review of Introduction to Linear Regression Analysisand Applied Linear Regression Models Journal of the American StatisticalAssociation 80 487ndash488

Applied Regression Analysis for Business and Eco-nomics (3rd ed) by Terry E Dielman Paci c GroveCA Duxbury 2001 ISBN 0 534-37955-9 viii C 647 pp$8995

As stated in the Preface this text is designed for a one-semester course inregression analysis for business and economics undergraduates and for MBAsI believe that this book meets this objective very well It presents regressionconcepts and techniques in a clear and concise format avoiding the use ofunnecessary mathematical rigor Assumptions behind a regression model are

presented along with the importance of model validation Applications ofregression analysis in the business setting are given with emphasis on toolsused to examine possible relationships between two or more variables Read-ers are assumed to have had an introductory course in statistics the topics ofwhich are covered in Chapter 2 Major changes in the third edition include

yuml updated datasets as well as additional problems and examples involvingreal data Some datasets come from actual business settings whereas othersare taken from the literature

yuml addition of Excel outputyuml Chapter 5 has been split into Chapters 5 and 6 The former discusses

tting curves to data and the later discusses assessing assumptions of theregression model

Chapters 2ndash10 each contains sections with actual examples and problemsets focusing mainly on business applications Additional exercises are pro-vided at the end of these chapters and answers are provided for all odd-numbered problems One of the bookrsquos strong features also included at theend of each chapter is illustration of the outputs of regression analysis resultswith different software packages including Minitab Excel and SAS Excelis included because it is prevalent throughout the business world The authorassumes that the reader has access to a computer and statistical software andemphasizes that using this software is an important part of the learning pro-cess Datasets for exercises in the text are available from Duxburyrsquos website(httpwwwduxburycom) in Excel Minitab SAS or SPSS formats

After a brief introduction to regression analysis in Chapter 1 basic statis-tical concepts are reviewed in Chapter 2 Such topics as descriptive statisticsdiscrete random variables and probability distributions the normal distribu-tion estimating a population mean hypothesis tests about a population meanand estimating the difference between two population means are covered

Chapter 3 presents various facets of simple regression analysis includingusing simple regression to describe linear relationships using regression as adescriptive technique making inferences from a simple regression analysisassessing the t of a regression line predicting or forecasting with a simplelinear regression equation and tting a linear trend to time series data Theauthor also discusses some important cautions when interpreting regressionresults such as association versus causality and forecasting outside the rangesof explanatory variables

The discussion of linear relationships is expanded to included multipleregression analysis in Chapter 4 Many of the same topics discussed in Chap-ter 3 are described for multiple regressions This chapter includes a luciddiscussion of choosing between two regression models full and reduced byusing separate regressions or by using conditional sums of squares

Chapter 5 expands on the earlier chapters to include tting curves to dataSuch topics as various transformations of the independen t variable and ttingcurvilinear trends are presented in good detail The former techniques may beused to minimize multicollinearity between various powers of the independentvariable Centering of the various powers of the independent variables mayalso avoid multicollinearity problems The latter technique is also useful for tting so-called S-curves used for modeling demand for products over theirlifetimes

In my opinion Chapter 6 is the most valuable (and lengthy) of the bookrsquos10 chapters Various ways of assessing assumptions of the regression modelare presented such as using residual plots for checking variance homogeneityaround the regression line Guidelines are offered for choosing which cor-rection to use for nonconstant variance such as natural log or square roottransformations of the dependen t variable The use of plots to assess whethererrors are normally distributed (particularly important for small sample sizes)is discussed Although corrections to violations of the normality assumptionsuch as BoxndashCox transformations are mentioned they are not elaborated onThe detection of outliers and leverage points (observations that may affectthe regression line) is discussed along with diagnostics such as Cookrsquos Dor DFITS statistics To determine whether errors are autocorrelated the useof the DurbinndashWatson or Durbin h test is described A good discussion ofcorrection for rst-order autocorrelation is presented I found one error inthe text the correct decision is to not reject the null hypothesis that rst-order autocorrelation is not a problem in Example 612 Finally an excellentdiscussion of consequences and detection of multicollinearity is presentedAlthough the author does not discuss remedial measures for multicollinearityin any detail he does give a reference for learning about such techniques asridge or principal components regression The author does point out however


BOOK REVIEWS 193

that if a model is used for forecasting then corrections to multicollinearityare probably unnecessary

Chapter 7 discusses the use of indicator or dummy variables for incorpo-rating qualitative information into regression analysis This qualitative infor-mation can represent two or more categories Seasonal patterns can also bemodeled in a regression equation by using indicator variables For exampletrends on a quarter-by-quarter basis can be modeled using three indicatorvariables in a regression equation The use of interaction variables useful fortesting for differences between the slopes of two regression lines is presented

Various variable selection procedures are discussed and compared in Chap-ter 8 If an important variable is omitted from a regression then estimatesof the coef cients in the regression equation become biased However if anunimportant variable is included then the standard errors of the coef cientsbecome in ated making inferential results less dependable Given a responseplus a set of explanatory variables one could examine all possible regressionequations using summary statistics of each to adopt a best regression Theauthor suggests using either R2 (adjusted or unadjusted) or Cp which mea-sures the total mean squared error of the tted values of the regression Thelatter statistic takes into account both random error and bias But a drawbackof using the all possible regressions approach could be computational cost ifthe number of explanatory variables is large

Other variable selection techniques such as backward elimination forwardselection and stepwise regression are also discussed Of these three the for-ward selection approach is generally considered the least reliable While usingany of these variable selection techniques knowledge of the subject matterbeing studied as well as the proper use of residuals plots and diagnostic statis-tics are vital in choosing the proper model One-way and two-way analysis ofvariance (ANOVA) as well as ANOVA using a randomized block design arepresented in Chapter 9 If blocks are the subjects to be used in an experiment(example on p 526 subjects are salespeople) then this design is referred toas a repeated measures design Analysis of covariance (ANCOVA) is onlybrie y mentioned without an example This procedure is used to increaseprecise measurement of treatment effects as discussed with many examplesof interest to the business community by Lunneborg (1994)

The last chapter is an introduction to discriminant analysis and logisticregression The latter approach is useful for modeling probabilities of a two-level categorical (eg passfail) dependent variable because the output isrestricted to be between 0 and 1 I was disappointed that the author doesnot describe the logistic regression technique in more detail with more exam-ples of not only two-level but also of multilevel categorical variables Thisapproach has many potential applications to business and economics (seeLunneborg 1994 for examples)

Overall I found this text to be well written with many examples relevantto the business world The statistical concepts in each chapter are clearlyand cogently explained and are important to anyone who relies on regressionoutputs to make business decisions Although I do not recommend this as apreferred general text above the plethora of other excellent texts in regressionanalysis I do recommend it for its intended audience

Henry W Altland

Corning Inc

REFERENCE

Lunneborg C E (1994) Modeling Experimental and Observational DataBelmont CA Duxbury

Eliciting and Analyzing Expert Judgment A PracticalGuide by Mary A Meyer and Jane M Booker Philadel-phia 2001 ISBN 0-89871-474-5 xxxii C 459 pp $85

This book covers everything one could possibly use or want to know aboutsimple but reliable procedures for eliciting and analyzing expert judgmentThis SIAM 2001 edition is an unabridged republication of the work rstpublished by Academic Press in 1991 The authorsrsquo intent is for the book to bea usable and readable guide to help the layperson collect and analyze expertjudgment scienti cally Therefore it requires no special technical backgroundother than some basic statistical knowledge obtainable in any introductorystatistics course The authors write in a clear style and their explanationsfor some of the statistical analysis procedures are the best I have seen I amparticularly impressed with their clear explanations of Bayesian analysis

The book is intended as a practical guide or handbook which means onedoes not have to read it from cover to cover before collecting and analyzingexpert data However the authors do suggest scanning the book to becomefamiliar with the topics and structure and then consulting those sections at theappropriate point in onersquos investigations On page 15 the authors provide ahandy owchart for using the book as a reference for eliciting and analyzingexpert information The chapters are designed to be mostly self-containedwhich allows the reader to access information quickly and to enter at anypoint

The book is organized in three parts ldquo Introduction to Expert JudgmentrdquoldquoElicitation Proceduresrdquo and ldquoAnalysis Proceduresrdquo There are also fourappendixes each listing Fortran codes for recommended analyses andsimulation procedures along with a comprehensive clearly written glossaryexplaining important terms used in the text

Chapter 1 de nes expert judgment as informed opinion on a technicalproblem based on an expertrsquos training and experience The authors presentan overview of how expert judgment is elicited They make the point thatthey are interested not only in the answers to technical questions but alsoin problem-solving procedures Both of these topics should be pursued toconstruct a framework for a stable knowledge system

Chapter 2 addresses the common questions and pitfalls concerning expertjudgment that arise due to misconceptions about expert judgment The ldquoPit-fallsrdquo section discusses aspects of elicitation and analysis that have causedproblems in the past Typical pitfalls are the introduction of bias the limitednumber of items (seven) that experts can mentally juggle the desired ldquogran-ularityrdquo of the data for analysis and the ldquocondition effectrdquo on gathering andanalyzing data The authors claim that experts are not naturally Bayesian intheir orientation and that experts may not signi cantly or consistently performbetter than nonexperts

Chapter 3 discusses in general terms the effects of bias on human problem-solving processes Trying to understand the causes of bias requires that theresearcher become aware of how people solve problems A six-step programis suggested for handling biases

Part II (Chaps 4ndash10) discusses elicitation procedures Chapter 4 addressesproblems in selecting general questions and identifying speci c questions

Chapter 5 discusses ways of re ning the questions selected This chapteralso includes suggestions for presenting information required to understandthe question that is the background de nitions and assumptions for the ques-tion along with suggestions for ordering the information and for dividing thequestions into more easily understood parts

Chapter 6 addresses the task of selecting and motivating experts for twotypes of applications those meant to gather the expertsrsquo answers and thosemeant to gather data on the expertrsquos problem-solving process In additionthe authors maintain that the researcher must distinguish between substantiveexpertise and normative expertise and that two different approaches are neededto obtain expert information Substantive expertise comes from the expertrsquosexperience in the eld in question Normative expertise is knowledge relatedto the use of the response mode that is familiarity with probabilities oddscontinuous scales and so on This distinction is important when selectingexperts

Chapter 7 discusses six basic components of elicitation and gives advice onselecting the most appropriate methods within these components Chapter 8discusses how the elicitation methods selected in Chapter 7 can be tailored toa researcherrsquos speci c application

Chapter 9 discusses practicing elicitation and training project personnelThese activities provide the last check on all of the aspects of elicitationdesign before information is actually elicited Project personnel can identifyany remaining problems by practicing and testing different parts of the elici-tation process

Chapter 10 discusses in detail how to schedule prepare for and conductthe elicitation Three techniques for obtaining expert data are covered Infor-mation is provided for monitoring and adjusting for bias during the elicitation

Part III is concerned with the procedures for analyzing an expertrsquosresponses Chapter 11 introduces the techniques for analyzing expert dataThis chapter discusses letters and symbols (ie notational conventions) basic statistical concepts and speci c analysis techniques includingsimulation techniques the Bayesian method and approaches and dataanalysis techniques The simulation methods discussed are Monte Carloand bootstrap Data analysis includes multivariate analysis correlationgeneral linear models cluster analysis factor analysis discriminate analysisanalysis of variance and Saatyrsquos method (the decision-analytic tool) An


194 BOOK REVIEWS

excellent discussion and explanation of Bayesian techniques is given onpages 230ndash233

Chapter 12 discusses initial procedures for looking at data Topics coveredinclude the granularity of the data and its possible conditional structure Theauthors recommend quantifying as much information as possible and suggestways to do this This quanti ed information should be placed in the databasefor further analysis

Chapter 13 explores ways of understanding the resulting database structureIt provides methods for exploring the relationships between ancillary variablesand answer-response variables and also ways of exploring relationships withineach of these variable sets The results of these analyses are not meant to beused as the nal results within the investigation but rather are preliminaryto further statistical investigation Correlation factor analysis cluster analy-sis and general linear models are some of the techniques used to explorerelationships between variables

Chapter 14 discusses correlation and how it relates to the application ofexpert judgment The relationship of correlation to bias is also covered with asubsection devoted to a detailed description of a 14-step method for detectingcorrelation among experts

Chapters 12ndash14 are concerned with investigating and conducting a prelim-inary analysis of data for the purpose of becoming familiar with informationgathered from the elicitation Chapters 15 and 16 focus on nal analysis pro-cedures that establish interpretable conclusions Chapter 15 presents modelingtechniques that may yield inferences and also suggests ways for describingexpertsrsquo answers and terms of the variables General linear models are empha-sized Multivariate methods such as factor discriminate and cluster analysisare used to appropriately model the multivariate structure of the database Butthe authors do not recommend using these techniques for the nal conclusionsbecause of the assumptions required to use them They recommend moreapplicable modeling techniques based on decision analysis methods whichcan be described as conditional models

Chapter 16 discusses combining schemes for aggregating the responsesof experts and the environments in which the schemes would be used Italso discusses the relationship of the aggregation problem to the problem ofcharacterizing uncertainties Particular emphasis is placed on Saatyrsquos methodof weight determinations

Chapter 17 is concerned with characterizing uncertainties It discusses thefour basic sources of uncertainty and reviews the procedures for obtain-ing uncertainty measures modeling uncertainties and comparisons of themethods

Chapter 18 the bookrsquos nal chapter is concerned with making inferencesfrom the data Two types of inferences are discussed general and statisticalThe authors emphasize that expert information does not allow one to makeinferences beyond the available knowledge base They state that the infer-ences made do not necessarily represent a true state of nature nor are theystatistically based inferences

This book belongs on the shelf of every researcher who is using or think-ing of using expert information in their work It provides a comprehensiveguide to simple but appropriate ways to elicit expert data as well as how toanalyze it

J Charles Kerkering

National Institute of Occupational Safety and Health

Fitting Statistical Distributions The GeneralizedLambda Distribution and Generalized BootstrapMethods by Zaven A Karian and Edward J DudewiczBoca Raton FL CRC Press 2000 ISBN 1-58488-069-4 xvii C 438 pp $7995

The generalized lambda family of distributions is a very broad familyof continuous univariate probability distributions The authors have been atthe forefront in investigating this distribution In this book they thoroughlyexplore the relationship of the generalized lambda family of distributions tomany commonly used families of distributions A main objective is to demon-strate that some member of the generalized lambda family can be found toclosely approximate most well-known distributions The major emphasis is onproviding body of techniques for tting the generalized lambda distributionto data

Chapter 1 provides an introduction to the generalized lambda family ofdistributions The multitude of shapes of the generalized lambda family of

distributions are explored in detail Chapter 2 discusses tting the generalizedlambda distribution to data The methodology is based on method-of-momentestimators using the rst four sample moments The authors give an extensivediscussion of the region in the skewness-kurtosi s plane that is covered by thegeneralized lambda family of distributions They also look at approximatingsome well-known distributions using generalized lambda distributions havingthe same rst four moments Chapter 3 develops a generalized beta distribu-tion and uses it to t data whose skewness and kurtosis are not in the regioncovered by generalized lambda family of distributions Chapter 4 uses per-centiles to t the generalized lambda distribution to data Chapter 5 presentsa bivariate generalized lambda family of distributions The authors discussapproximating some well-known bivariate distributions using the bivariategeneralized lambda family of distributions They also develop an algorithmfor tting the bivariate generalized lambda distribution to data Chapter 6discusses the use of a generalized bootstrap to obtain con dence intervals forfunctions of the distribution for the generalized lambda family of distribu-tions Programs (in Maple) for tting the distributions to data and tables forobtaining ts of the generalized lambda distribution to data are provided inappendixes

The authors provide a thorough exploration of the generalized lambda fam-ily of distributions and its use in the tting of data Practitioners who wishto t data with a generalized lambda distribution will nd this book usefulNumerous examples with actual datasets illustrate the utility of the techniquesHowever practitioners who want to compare a variety of methods for ttingdistributions to data will be disappointed in the bookrsquos narrow focus Theamount of detail needed in the exploration of the generalized lambda familyof distributions precludes the authors from providing out a more detaileddiscussion of other methods of tting distributions to data

Statisticians may complain about the lack of theoretical results concerningthe techniques in the book For instance no results concerning the ef ciencyof estimators are given The theoretical properties of estimators are not devel-oped The generalized lambda family of distributions is used solely as a broadfamily of distributions for the empirical tting of data There are no physicalmechanisms that lead to this family Also interpretations for the parametersare not provided

The authors are dismissive of the bootstrap (Chap 6) arguing that it isnot useful in estimating the distribution of statistics dealing with the extremesof a set of data This is well known in the literature on the bootstrap (seeeg Efron and Tibshirani 1993 p 81) However the bootstrap does wellin handling statistics that deal with the center of the distribution such asthe sample mean and is a valuable tool for estimating the standard error ofthese statistics The authors develop a ldquogeneralized bootstraprdquo and apply it tothe generalized lambda family of distributions Other authors have called thisprocedure the ldquoparametric bootstraprdquo (see eg Efron and Tibshirani 1993p 53) and have applied it successfully to other distributions

In summary the authors have presented a complete exploration of the useof a particular family of distributions in tting data This book may be toospecialized for the general practitioner but could be of use to practitionerswho want to use a general family of distributions to t to data

Thomas E Wehrly

Texas AampM University

REFERENCE

Efron B and Tibshirani R J (1993) An Introduction to the BootstrapNew York Chapman and Hall

Time-Series Forecasting by Chris Chat eld BocaRaton FL Chapman and HallCRC 2001 ISBN 1-58488-063-5 xi C 267 pp $6995

This well-written and comprehensive review of current time-series andforecasting methods should quickly earn a place among standard referencematerials The author states that the book ldquois primarily intended as a referencesource for practitioners and researchers in forecasting who could for exam-ple be statisticians econometricians operational researchers managementscientists or decision scientistsrdquo (p ix) Hence this book focuses on gen-eral methods especially for economics government industry and commercebut ignores specialized areas such as meteorology or judgment forecastingIt presents these methods from a utilitarian perspective clearly explaining


BOOK REVIEWS 195

what these methods may potentially accomplish and what risks they entailBrief summaries explain the related theory in plain prose Numerous ref-erences direct the interested reader to more information on speci c detailsand tangents theoretical results and special applications In addition thedata for four real world applications can be found on the authorrsquos website(httpwwwbathacuk mascc)

This book covers such a broad range of topics in time series and forecastingthat it is dif cult to summarize its contents in a meaningful way Thus I havelisted the main topics of each chapter and highlighted especially useful andinteresting discussions by the author on particular topics

Chapter 1 gives an overview of the bookrsquos scope It addresses the crucialdifferences between ldquomethodrdquo and ldquomodelrdquo and between univariate and mul-tivariate methods It also discusses the problems of judgment forecasts theimportance of formulating a problem carefully and the dangers of extrap-olation inherent in forecasting Chapter 2 reviews the basic terminology oftime series analysis The many topics covered include measurement scalesobjectives of time series analyses seasonal variation and trend basic graphicalmethods and transformations of data stochastic process theory autocorrela-tion and autocovariance the classi cation of univariate models [eg randomwalks autoregressive (AR) models moving averages (MA)] and correlo-grams

Chapter 3 covers an extensive range of univariate models It begins withautoregressive integrated moving average (ARIMA) models (and their parts)and the related issues of seasonality periodicity fractional models and unitroots It continues with state-space models growth curve models nonlinearmodels (eg ARCH and GARCH models) and provides a thoughtfully skep-tical discussion of the ldquohot topicsrdquo of neural nets (NNs) and chaos Section 35focuses on model building selection and checking and includes a helpful listof guiding concepts (p 80)

Chapter 4 focuses on univariate forecasting methods the heart of the bookIt rst describes the general problem of prediction and loss functions thencompares and contrasts model-based forecasting methods [eg BoxndashJenkins(ARIMA) procedure Kalman lters nonlinear models] and ad hoc forecastingmethods (eg HoltndashWinters procedure combinations of forecasts)

Chapter 5 covers multivariate forecasting methods and discusses why theyare so much harder than univariate methods The concepts covered hereinclude feedback open- and closed-loop systems the problem of out-of-sample forecasts leading indicators and multivariate cross-correlations Itthen presents a strong argument that regression models are generally inap-propriate for time series problems and uses this discussion to motivate thedevelopmen t of alternative models (p 117) These alternatives include transferfunction models vector versions of ARIMA models error-correction models(based on cointegration) and econometric models among others

Chapters 6ndash8 shift from a consideration of models and methods to the goalsof forecasting These chapters would be well worth reading for experts in theaforementioned methods to gain from the authorrsquos broad perspective

Chapter 6 gives a comparative assessment of forecasting methods startingwith a list of practical criteria for choosing the ldquobestrdquo forecast It also consid-ers measures of forecast accuracy (eg PMSE MAE MAPE) and discusseslessons learned from forecasting competitions and case studies Finally itgives advice on strategies for choosing a forecasting model including severalhelpful lists of criteria (pp 170 173 and 178)

Next Chapter 7 discusses the challenges of interval forecasting It exten-sively reviews methods for constructing prediction intervals (eg modelapproximate empirical simulation and Bayesian-based methods) It thencompares the results of different methods explains why prediction intervalstend to be too narrow in practice and concludes with some practical recom-mendations for constructing prediction intervals (p 213) Chapter 8 discussesthe many sources of model uncertainty and their impact on forecast accuracyIt also examines the balance between model building and data dredging andillustrates this with an example of NN models applied to BoxndashJenkinsrsquo clas-sical airline data It ends with cautionary advice about model selection and adiscussion of practical methods for dealing with model uncertainty Finallythe book contains an extensive list of references with notes as to where eachreference is cited

One of the bookrsquos strengths is that after presenting a topic the authorroutinely brings his personal views and experiences into the picture Anotherstrength is the numerous checklists of ideas throughout which serve to clarifyconcepts and reinforce key points that are easy to forget The authorrsquos advicecomes across as thoughtful guidance and makes this book more interestingto read

Nonetheless I wish that this book had included a few more real-worldapplications and that it had developed those that are presented more fully Tosome extent the limited number of such examples and the briefness of detailre ect the nature of reference books Nevertheless given the authorrsquos clearsense of the need to analyze problems in terms of their applied context Ithink that the foregoing suggestion would have enhanced the presentation ofhis perspective especially in the nal chapters

In summary this book represents a helpful and enlightening reference forpracticing statisticians (among others) who work with time series and fore-casting applications and who wish to think critically about current practicein these areas This book could also be the core text of a graduate seminaron forecasting for students with a good background in time series analysisHopefully some bright and ambitious students will be stimulated to tacklethe diverse open questions and statistical challenges that the author raises

Craig B Borkowf

National Cancer Institute

Practical Time Series by Gareth Janacek LondonArnold 2001 ISBN 0-340-71999-0 xv C 156 pp $3495

This book is intended for a nonspecialist audience It may be used as atext for an advanced undergraduat e or a masterrsquos-level time series course It isequally suited for a practicing statistician working in industry or government The book assumes knowledge of undergraduat e calculus and basic statisticalmethods and probability The coverage is pretty broad and the approach isapplication oriented mostly intuitive and hands-on problem solving withouttoo much mathematical detail The examples are carefully selected and wellpresented The author has used R freeware (an S-PLUS clone) for all of theexamples in the book However the presentation is software independent and any standard statistical package may be used to solve the problems inexercises The data and codes used in the book are available free of chargefrom the authorrsquos website

The book starts with Chapter 1 introducing smoothing as a tool for explor-ing the data Many of the discussions are ad hoc in nature which is inline with the booksrsquos objective to address a nonspecialist audience The non-specialist user will nd such excellent not-too-theoretica l discussions veryuseful for his or her practical time series problems The chapter introduces thetopics of moving averages regression smoothers median smoothers and itsre nements LOWESS k nearest-neighbor and kernel smoothers Chapter 2discusses exponential smoothing and generalized exponentia l smoothing indetail

In Chapter 3 Janacek shifts attention from an ad hoc approach to a moreformaltheoretical approach However the discussion never becomes toomathematically complicated Janacek introduces just the right amount oftheoretical details needed to deliver the essential message to the bookrsquospractitioner audience He also introduces Woldrsquos theorem without gettinginto the proof The reader will be able to appreciate the implications ofthe result and apply it in prediction problems The ARMA and ARIMAmodels together with YulendashWalker equations and the topics of estimation andprediction are discussed in detail Janacek nicely presents the concepts ofinvertibility and stationarity and its implications without going into dif cultmathematics

Chapter 4 is devoted to the general state-space approach and Kalman lterIndeed the likelihood estimation routine used in the book is based on thisapproach Janacek also demonstrates why such modeling is often preferablein practical problem solving

Chapter 5 deals with the important topics of estimation identi cationmodel comparison and model selection Both informal techniques and formalcriteria for model selection such as the Akaike Information Criterion (AIC)and the Bayes Information Criterion (BIC) are presented The chapter con-cludes with an example of tting a SAR IMA model

The book strikes a good balance between the time domain and the fre-quency domain approaches In Chapters 6 7 and part of Chapter 8 Janacekshows why in many situations the frequency domain approach is the naturalchoice over time domain techniques This is particularly true for engineeringapplications However for rst-time readers the frequency domain approachis often a little more dif cult than the time domain approach As opposed tothe previous chapters in certain sections of Chapter 6 the reader may some-times get lost particularly studying the book on his or her own rather than


196 BOOK REVIEWS

as a student in a class taught by a professor Janacek probably should havegiven a little more explanation of some of the involved concepts of Chapter 6a little less hand-waving and a little more in-depth discussion would havebeen helpful The linear lter lter design aliasing periodogram DiscreteFourier Transform Fast Fourier Transform estimation of spectrum are themain topics of discussion in Chapters 6 and 7 Chapter 8 is devoted to theanalysis of multiple correlated time series a very important concept from apractical standpoint The book ends with Chapter 9 on R language

Overall Janacek has done a good job In most cases the intuitive hands-onapproach without mathematical proofs increases the bookrsquos readability How-ever I should mention that the book contains several printing errors Some-times these mistakes are very frustrating and the reader particularly one notfamiliar with the topic of time series should read very carefully Signi canteditorial effort should be devoted to correcting these errors in the bookrsquos nextedition

Pradipta Sarkar

United Technologies Research Center

Practical Time-Frequency Analysis Gabor andWavelet Transforms With an Implementation in S byReneacute Carmona Wen-Liang Hwang and Bruno TorreacutesaniSan Diego Academic Press 1998 ISBN 0-12-160170-6 490 pp $65

The purpose of this book is to provide a review of techniques in timefrequencytime scale and to perform analyses using S language Statisticaldescriptions applications and limitations for each technique are given Statis-ticians in particular will nd many useful tools to detect denoise and recon-struct signals and most importantly to perform spectral analyses of nonsta-tionary signals The book is organized in four parts I background and anintroduction in time frequency and time-frequency and the spectral theoryof stationary random process II Gabor and wavelet transforms continuousGabor and wavelet transformation discrete time-frequency transforms andalgorithms III signal-processing applications and IV Swave library for theS language to run examples mentioned earlier in the text

Each chapter includes theory techniquesmethods examples along withcorresponding gures and the S language commands needed for each tech-nique Each chapter ends with a summary note and additional references of awider range of related materials

Using window sizes to discriminate between harmonic components isexplained in Chapter 3 Nonparametric spectral estimation techniques forlocal behavior characterization and continuous-discret e Gabor transformation(CGT) with Gaussian windows are used for different datasets Continuousand discrete wavelet transformation is introduced as an alternative method toGabor time-frequency transformation for performing local time scale analysis(Chap 4)

Chapter 6 introduces a signal-processing application on data with stationaryincrements and especially the processes of fractional Brownian motionTechniques such as the wavelet spectral estimate are found to be morerobust and better predictors of the Hurst exponent than the periodogramChapter 7 deals with the analysis of frequency-modulate d signals using theridge of wavelet transform curve in localization properties of asymptoticsignals in time frequency domain Differential and integrated methods indetermining local amplitude frequencies with unique or several components and pure or noisy signals are also introduced The authors present differentmethods to reveal the salient features of signals from their time frequencytransform

Chapter 8 presents reconstruction of signals from a known signal featuringldquodenoisingrdquo using threshold methods Several methods are used includingnonparametri c regression including coef cient thresholding regression andthresholding wavelet coef cients the smoothing spline and reconstructionfrom extrema of the dyadic wavelet transform Included are detailed descrip-tions of testing for the presence of noisy and transient noise using the wavelettransform maximal information and trimming the maxima in an effort toapproximate the original signal lead to the easily reconstructed correspond-ing component The chapter also covers the developmen t of a smoothingspline-based method for signal reconstruction from ridges with different cases(eg wavelet transform) stating the problem choosing a functional form andpenalty are all demonstrated through data-laden examples

The last chapter explains downloading and installing the Swave packageand Swave S function A rich bibliography is divided into three sectionsGeneral Reference Wavelet Books and S-PLUS Books This book providesa good reference for theory and examples of spectral analysis for statisticiansand nonstatisticians alike

Maliha S Nash

US Environmental Protection Agency

Geometric Data Analysis An Empirical Approach toDimensionality Reduction and the Study of Patterns byMichael Kirby New York John Wiley 2001 ISBN 0-471-23929-1 xvii C 363 pp $6495

This book discusses one of the key issues in the information industrydimension reduction The appeal of the title is much stronger to people whowork in such elds as data processing data modeling and data mining I feelsomewhat disappointed however this is a purely mathematical book Thisdisappointment is not because of my prejudice as a statistician but ratherbecause the book does not address issues about complicated variations fromhigh-dimensiona l data which are crucial in many real applications The authorclaims that ldquoapproaches for the analysis of patterns can be divided into twogroups probabilistic and deterministicrdquo (p 24) and the book deals ldquoalmostexclusively with techniques that fall into the latter categoryrdquo (p 24) As amathematical text it presents a nice treatment for constructing dimensionality-reducing mappings both linear and nonlinear Although more than 90 ofthe material in the book will be familiar to most readers this is one of the rst books to put a variety of methods together and to elaborate from thegeometric angle For statisticians although the book provides a valuable sum-mary of mathematical formulations for many different methods it falls shorton guidelines for a wide range of applications in dimension reduction

The book comprises four parts ldquo Introductionrdquo ldquoOptimal Orthogonal Pat-tern Representationsrdquo ldquoTime Frequency and Scale Analysisrdquo and ldquoAdaptiveNonlinear Mappingsrdquo Part 1 serves as the prerequisite and presents the singu-lar value decomposition (SVD) in detail SVD plays a fundamental role in theconstruction of a mapping in the linear space from high to low dimension thebookrsquos theme Part 2 discusses the KarhunenndashLoegraveve (KL) expansion whichis the foundation of the book The author devotes two chapters 3 and 4to the KL transform and its variations I believe that most statisticians arefamiliar with the KL transform because it is nothing more than the principlecomponents analysis Chapter 4 touches brie y on implementation of the KLtransform when missing data are present Although an iterative algorithm forestimating a KL basis with missing data is given in Section 41 statisticiansmay nd that this chapter suffers from a lack of discussion on the stabilityand reliability of the algorithm In statistics there is a relatively large bodyof literature on dealing with missing data including a rich content on theEM algorithm approach The reason why I believe the author did not addresssuch an approach is not because of the geometric approach but because ofthe nonexistence of statistical concepts Chapter 4 also discusses to the exten-sion of the KL transform to the Hilbert space Clearly this is an interestingperspective in mathematics but it is not a major concern in statistics

Part 3 focuses on Fourier analysis including some theories on the wavelettransformation Fourier analysis ts naturally into this book because theFourier coef cients form a basis for an inner product space One of its directapplications wavelet transformations is discussed in Chapter 6 Some proper-ties of wavelets such as multiresolution analysis are concisely explained Thediscussions in Chapter 6 serve well as an introduction of the wavelet theorybut the reader will de nitely need to look elsewhere for more advanced topicson wavelets

Part 4 contains three chapters 7 ldquoRadial Basis Functionsrdquo 8 ldquoNeural Net-worksrdquo and 9 ldquoNonlinear Reduction Architecturerdquo Radial basis functions areusually considered part of research on neural networks and are covered inmost neural network texts Assuming that a radial basis function is an activa-tion function in neural networks the output function as a nonlinear mappingis basically a radial basis function expansion Chapter 7 focuses on a simplerversion of a radial basis function with a given shape parameter With such anassumption the author is able to convert the problem to the least squares set-ting Therefore many algorithms are available for solving the problem someof the commonly used ones are illustrated in Sections 73 and 74 Chapter 8discusses feed-forward networks only and mainly covers the backpropagatio n


BOOK REVIEWS 197

algorithm The author shows that as a class of nonlinear mappings betweenvector spaces neural networks (including radial basis functions) can indeedbe a useful method of dimension reduction

Chapter 9 is as the author claims ldquothe basic goal of the book that is thedevelopmen t of methods for constructing empirical dimensionality-reducin gmappings of datardquo (p 299) The so-called WRN (Whitney reduction network)based on Whitneyrsquos embedding theorem as a global approach is covered indetail However in these global approaches as the author notes ldquowhen thedomain dimension is large are far less practicalrdquo The local approach is theTaylor expansion of the nonlinear mapping Thus the linear term of the Taylorexpansion can be applied by the KL procedure a local KL algorithm is alsogiven (p 326)

This book provides a valuable summary of data reduction Of course math-ematical theory is an important part of the foundation for data analysis Butdue to the limitation of the bookrsquos purely mathematical viewpoint only theleast squaresndashtype criteria are discussed for determining an optimal basisMany other approaches including maximum likelihood are not mentionedat all This view limits the bookrsquos applicability

Yachen Lin

First North American National Bank

Introduction to Graphical Modelling (2nd ed) by DavidEdwards New York Springer-Verlag 2000 ISBN 0-387-95054-0 xv C 333 pp $6995

Graphical modeling is a form of multivariate analysis that uses graphs torepresent models with conditional independence as the fundamental model-building criterion Graphical models are an attractive alternative to moretraditional models because they deal well with mixed models have a solidtheoretical basis and have natural ldquographicalrdquo graph depictions that aid intu-ition and invite critical thought The rst edition of this book published in1995 provided an excellent practical companion to more theoretical worksuch as that of Whittaker (1991) The second edition adds approximately60 pages devoted mostly to new chapters on causality and modeling usingdirected graphs Both of these chapters are very welcome indeed my maincomplaint about the rst edition was the lack of information on directedgraphs

The bookrsquos emphasis is rmly on model structure both in organizationand in discussion The book does not give standard proofs or provide imple-mentation details for those wishing to implement software themselves itsgoal is to introduce the concepts and to show how these concepts may beapplied The book starts with a discussion of core ideas and then introducesloglinear models and graphical Gaussian models before its longest chapteron mixed models The remaining chapters deal with hypothesis testing andmodel selection for mixed models and the aforementioned topics of causalityand directed graphs

This bookrsquos strength is its accessibility Numerous illustrations and exampledatasets are well integrated with the text and the author provides scripts fora program MIM that can be used to replicate experimental results and tryalternatives A version of M IM is freely downloadable from the Internet andis easy to install and use The examples are well chosen I was particularlypleased that the author clearly treated the datasets as interesting in their ownright not simply as a foil for demonstrating techniques

In summary Edwards presents a clear engaging introduction to graphicalmodeling that is very suitable as a rst text and should stimulate readers toexplore and use this methodology for their own data

Graham J Wills

Bell Labs Lucent Technology

REFERENCE

Whittaker J (1991) Graphical Models in Applied Multivariate StatisticsChichester UK Wiley


This section reports on new editions of books previously reviewed in Tech-nometrics collections of papers and conference proceedings and other statis-tics books that should have some interest for the readership Selections andcomments do not represent any perspective of the editorrsquos employer or of thesponsoring societies

Eric R Ziegel

BP

Engineering Statistics (2nd ed) by Douglas C Mont-

gomery George C Runger and Norma F Hubele New York Wiley 2001 ISBN 0-471-38879-3 xiv C 494pp $9395

This is the ldquoliterdquo version of Applied Statistics and Probability for Engineers(Montgomery and Runger 1999) hereafter called MR reported by Ziegel (inpress) This second edition (2E) of the condensed version of MR has beenpublished like its rst edition (1E) a couple of years after the publicationof MR See Ziegel (1999) for a thorough report on the 1E of the condensedversion for the rst edition of MR (MR1)

An interesting aspect of the relationship between this book and MR is thatthe reorganization of some of the material from MR1 that was done to createthe 1E of this book was subsequently applied by the authors in preparingMR Because this book had its gestation as a revision of another book thereperhaps was little need to further revise it The table of contents is unchangedand the overall result of the revision is only 20 or so additional text pages

The description of the contents of 1E by Ziegel (1990) suf ces also forthe 2E Ziegel (2002) did not feel that owners of MR1 needed to buy MRowners of the 1E similarly do not need the 2E The publisher has actuallyreduced the price so the 2E costs less than its much larger bigger brotherMR a situation that did not initially exist with the 1E and MR1

REFERENCES


Ziegel E (1999) Editorrsquos Report for Engineering Statistics by D Mont-gomery G Runger and N Hubele Technometrics 41 271

(2002) Editorrsquos Report for Applied Statistics and Probability forEngineers (2nd ed) by D Montgomery and G Runger Technometrics44 93

Fault Detection and Diagnosis in Industrial Systemsby L Chiang E Russell and R Braatz New YorkSpringer-Verlag 2001 ISBN 1-85233-327-8 xiv C 279pp $42

This book is part of the publisherrsquos softcover textbook series on AdvancedTextbooks in Control and Signal Processing See Ziegel (2002) for a reviewof Fortuna et al (2001) Fault detection and diagnosis are hot topics in controlengineering that are mostly subsumed by statistical methods for multivariatestatistical process control (MSPC) As the recent article by Kano HasebeHashimoto and Ohno (2001) demonstrates generally the methodology is pro-ceeding at a rapid pace in the chemical engineering literature without anyhelp from statisticians Likewise as another recent paper by Amand Heyenand Kalitventzeff (2001) shows fault detection methods similarly are beingcreated almost exclusively by the engineers despite the ostensibly statisticalnature of the problem

The authorsrsquo goal was to write ldquoa single textbook that covers data-drivenanalytical and knowledge-base d process monitoring methodsrdquo (p viii) Thesedifferent types of methods are all described in the rst chapter About two-thirds of the book is devoted to the data-driven methods The book concludeswith single chapters on analytical methods and knowledge-base d tools Theformer methodology requires complex mathematical models built on engineer-ing principals and monitored by examining the residuals which can be eitherdifferences between the predictions and the data or parameter differences from


198 BOOK REVIEWS

on-line model updates If the rst principles provide an accurate represen-tation then they provide useful additional information beyond just the dataAlthough using analytical methods entails parameter estimation and residualsdetermination the chapter is written in the heavily mathematical languageof control engineers without any of the statisticianrsquos concern for variabilityand statistical signi cance The chapter on knowledge-base d methods focusesprimarily on fault diagnosis through the use of causal analysis and expertsystems It also presents the pattern recognition procedures based on neuralnetworks or fuzzy logic which can be used to do fault diagnosis

The chapters on data-driven procedures are divided into three sectionsbackground methods and applications The background chapters present amathematical look at T2 statistics and discriminant analysis as the basis forfault detection and diagnosis The section on methods has chapters on princi-pal components linear discriminant analysis partial least squares and canon-ical variates analysis Compared to the presentations in a good multivariateanalysis textbook such as the recent book by Krzanowski (2000) reportedfor Technometrics by Ziegel (2001) this book generally captures only themathematical part of the methodology The book is about computational toolsnot modeling It gives the methods that provide the means for calculatingvarious fault-detection statistics but includes no statistical analysis The pri-mary thrust of the applications is the comparison of the different methodsusing data from a well-known simulation model The faults are imposed onthe simulation and then detected and diagnosed in the comparative study

REFERENCES

Amand T Heyen G and Kalitventzeff B (2001) ldquoPlant Monitoring FaultDetection Synergy Between Data Reconciliation and Principal Compo-nents Analysisrdquo Computers and Chemical Engineering 25 501ndash507

Fortuna L Rizzotto G Lavorgna M Nunnari G Xibilia M andCaponetto R (2001) Soft Computing London Springer-Verlag

Kano M Hasebe S Hashimoto I and Ohno H (2001) ldquoA New Multi-variate Statistical Process Monitoring Method Using Principal ComponentAnalysisrdquo Computers and Chemical Engineering 25 1103ndash1113

Krzanowski W (2000) Principles of Multivariate Analysis (rev ed) NewYork Oxford University Press

Ziegel E (2001) Editorrsquos Report on Principles of Multivariate Analysis (reved) by W Krzanowski Technometrics 43 498

(2002) Editorrsquos Report for Soft Computing by L FortunaG Rizzotto M Lavorgna G Nunnari M G Xibilia and R CaponettoTechnometrics 44 forthcoming

Scan Statistics by Joseph Glaz Joseph Naus and SylvanWallenstein New York Springer-Verlag 2001 ISBN0-387-98819-X xv C 370 pp $8495

Doing a report on a recent collection of articles on scan statistics by Glazand Balakrishnan (1999) (Ziegel 2001) I noted that it was a subject areain statistics that had previously escaped my notice I also commented that Ibarely could discover what a scan statistic was other than readily ascertainingthat scan statistics seemed clearly to be in the domain of the mathematicalstatistician This book which has the same senior author as the earlier workdoes a great job of helping the novice practitioner understand the value ofscan statistics The rst part of the book is titled ldquoMethods and ApplicationsrdquoAlthough it represents only about 30 of the book it provides a nice overviewof the formulas for probability calculations access to useful tables and anintroduction to relevant computer programs The second part of the book istitled ldquoScan Distribution Theory and Its Developmentrdquo All of the material inboth parts of the book is surrounded by a plethora of applications

The authors state succinctly at the beginning of the Preface that ldquoscanstatistics arise naturally in the scanning of time and space looking for clustersof eventsrdquo Working at a site that has a possible brain tumor cluster I haveno dif culty envisioning applications Many of these are summarized in theintroductory chapter where the scan statistic for the brain tumor situationwould be the largest number of occurrences in a xed window of time Therealso are discrete scan statistics particularly in the analysis of DNA sequences and scan statistics in two dimensions which arise particularly for cosmicray data obtained from the search for signals from objects or from radiationsources in space Anyone who ever gambled or participated in sports shouldsee the relevance of this methodology to the ldquohot streaksrdquo that characterizethose pursuits

Chapter 2 ldquoRetrospective Scanning of Events Over Timerdquo deals with thoseevents that are possibly unlikely occurrences The scan statistics are the largestnumber of events in a xed time and the shortest period of time containing a xed event Anyone can surely think of a few applications Most of the chaptercomprises discussions about 10 of these Chapter 3 ldquoProspective Scanning ofEvents Over Timerdquo deals with events from the familiar Poisson process Theapplications concern the monitoring of future data Some are familiar enoughsuch as overloads of telephone lines or arrivals of customers for servicesThis chapter has 11 examples In Chapter 4 ldquoSuccess Scans in a Sequence ofTrialsrdquo the model is binomial and immediately many yes-or-no-event s thatcan be viewed as exceptional such as out-of-control for processes or home-run-or-not for Barry Bonds should come to mind Most of the nine examplesinvolve quality reliability and other familiar applications

The last of the six applications chapters ldquoHigher-Dimensional Scansrdquo andldquoScan Statistics in DNA and Protein Sequence Analysisrdquo are much morespecialized and generally outside the realm of industrial statistics Each ofthese chapters also is profusely illustrated with a various applications The 11chapters that make up the remaining 70 of the book provide another doseof the same type of reading available in the book of Glaz and Balakrishnan(1999) The authors continue to illustrate the material effectively The materialis obviously much more cohesive in a textbook format This book would bea good library addition for any statistician

REFERENCES

Glaz J and Balakrishnan N (eds) (1999) Scan Statistics and ApplicationsNew York Birkhaumluser

Ziegel E (2001) Editorrsquos Report for Scan Statistics and Applications editedby J Glaz and N Balakrishnan Technometrics 43 246ndash247

Advanced Linear Modeling (2nd ed) by RonaldChristensen New York Springer-Verlag 2001 ISBN0-387-95296-9 xiii C 398 pp $7995

The title of this book re ects the authorrsquos methodology for unifying multi-variate analysis time series analysis and spatial modeling as various reformu-lations of linear model theory Someone besides the author must have foundthis perspective intriguing because here is the book again unifying evenmore methodologies This second edition contains new chapters in which atleast some of the models from nonparametric regression and response surfaceoptimization are similarly viewed as types of linear modeling

In his review of the rst edition Niple (1992) noted that ldquoChristensenhas provided us with an interesting if challenging book that contains manyvaluable insightsrdquo He cautioned that the book ldquowill be unsatisfying to mostof its readers because it is based on the authorrsquos unconventiona l approach tolinear modelsrdquo The reviewer was referring to Christensen (1987) with thiscomment Niple provided a thorough description of the six chapters for the1E Signi cant additions to this material are a section on repeated measuresmodeling in the chapter on multivariate linear models (Chap 1) and a sectionon models for lattice data in the chapter on kriging models (Chap 6)

Chapter 7 ldquoNonparametri c Regressionrdquo begins with an approach basedon orthogonal series It also includes heteroscedastic models splines kernelregressions and regression trees as methods considered There is an exampleon voltage drops versus time for batteries Chapter 8 ldquoResponse SurfaceMaximizationrdquo is limited strictly to that aspect of response surface modelingTopics include linear approximating functions steepest ascent and quadraticresponse functions This chapter primarily deals with a four-factor study forsome unspeci ed response versus the levels of four chemicals The selectivityof the topics is very apparent from the content of these two chapters

The author changed the bookrsquos title ldquoto indicate that it contains much newmaterialrdquo (p v) The book has not become any more advanced than it wasbefore However it still seems as noted by Niple (1992) to be ldquounsuitableas a primary textbookrdquo which Niple felt might be ldquoa valuable addition to agraduate-leve l course in linear model theoryrdquo


BOOK REVIEWS 199

REFERENCES

Christensen R (1987) Plane Answers to Complex Questions The Theory ofLinear Models New York Springer-Verlag

(1991) Linear Models for Multivariate Time Series and SpatialData New York Springer-Verlag

Niple E (1992) Review of Linear Models for Multivariate Time Series andSpatial Data by R Christensen Technometrics 34 354

The Six Sigma Revolution by George Eckes New YorkWiley 2001 ISBN 0-471-38822-X xiv C 274 pp $2995

Some authors can irritate me with their approach to their presentation oftheir material Here the author has used four footnote references at the endof Chapter 1 three referring to web sites and the fourth to a General Electricannual report Thereafter not one single reference is made to any avail-able resources Subtitled ldquoHow General Electric and Others Turned ProcessInto Pro tsrdquo the book actually separates itself successfully from other SixSigma books by focusing exclusively on helping upper management effec-tively implement Six Sigma within their own companies However despitethe bookrsquos minimal technical details because it is not targeted to black beltsand green belts other ldquonuts-and-boltsrdquo books such as those by Breyfogle(1999) reviewed by Gardner (2000) and Breyfogle (2001) reported by Ziegel(2001) were not deemed worthy of mention Ziegel (in press) particularlyrecommended Pande Neuman and Cavanagh (2000) for its value towardefforts for Six Sigma implementation but seemingly the only thing that thisauthor values is his own ideas

All scathing criticisms aside this is otherwise a very nice book With itslow price the publisher has targeted it toward the mass business marketUnlike the ground-breakin g Six Sigma book by Harry and Schroeder (2000)which was reported by Alexander (2001) this book does not attempt to blowmanagers away with pages of hype about the wonder and success of SixSigma After explaining how the quest for the ultimate quality improvementprocess progressed to what has become Six Sigma the author provides eightsteps to creating the business infrastructure that will allow for the successfulimplementation of Six Sixma Chapter 3 sets forth the theory that customersprocesses and employees are the ingredients for pro ts

Chapters 4ndash9 provide the authorrsquos own version of the Six Sigma DMAICmethodology (including two chapters on analysis) Design focuses on teamscustomers and process mapping another version of those ingredients forpro t Measurement focuses on two aspects the plan for data collection andthe calculation of the baseline sigma level The former includes some briefdiscussion of sampling and the latter deals with bell curves and process capa-bility Both analysis chapters deal with aspects of root cause analysis Thestatistical tools for analysis do not progress much beyond simple plottingbut the latter chapter does contain much material on designing factorial andfractional factorial experiments Improvement concentrates mostly on over-coming resistance to the new ideas that are determined by analysis to be thesolutions The chapter on control presents a lot of information on strategy andalso several pages of discussion of control charts

The last chapter lists the authorrsquos 10 concerns for the success of Six Sigmainitiatives to avoid failure Concern 1 says that ldquothe key to Six Sigma isstatistics statistics statisticsrdquo (p 244) The author voices the complaint thatldquomany Six Sigma consultants are statisticiansrdquo who ldquocreate the concern thatyou must turn your organization over to statisticiansrdquo (p 245) Perhaps that iswhy BP limits its statistician population to one In concern 10 ldquoignoring themanagement of change in your organizationrdquo (p 262) the author promoteshis next book Making Six Sigma Last Managing in a Changing EnvironmentDo I need mention that the author is head of a consulting group

REFERENCES

Alexander M (2001) Review of Six Sigma by M Harry and R SchroederTechnometrics 43 370

Breyfogle F (1999) Implementing Six Sigma New York Wiley(2001) Managing Six Sigma New York Wiley

Gardner M (2000) Review of Implementing Six Sigma by F BreyfogleTechnometrics 42 309ndash310

Harry M and Schroeder R (2000) Six Sigma New York DoubledayPande P Neuman R and Cavanagh R (2000) The Six Sigma Way New

York McGraw-Hill

Ziegel E (2001) Editorrsquos Report for Managing Six Sigma by F BreyfogleTechnometrics 43 382

Review of The Six Sigma Way by P Pande R Neuman and RCavanagh Technometrics 44 forthcoming

Soft Computing by L Fortuna G RizzottoM Lavorgna G Nunnari M G Xibilia andR Caponetto London Springer-Verlag 2001 ISBN1-85233-308-1 xii C 267 pp C CD $4995

This bookrsquos title seems like a complete misnomer The book is all aboutfuzzy logic neural networks genetic algorithms and their combinations andextensions All of these topics have always been pretty hard for me to grasp atleast in the cursory examinations that are done for my quick reviews This isnot a statistics book It comes from a slick soft-cover series that the publisherproduces for process control and signal processing applications Reviews areforthcoming for two other books in the series by Chiang Russell and Bratz(2001) and Chonavel and Vaton (2001) Control engineers seem to be muchmore adept than statisticians at adapting to these new technologies

In the authorrsquos words (p vii) soft computing ldquoprovides algorithms that areable to value to reason and to discriminate rather than to just calculaterdquoMore speci cally ldquosoft computing is a methodology tending to fuse synergis-tically the different aspects of fuzzy logic neural networks evolutionary algo-rithms and nonlinear distributed systems in such ways to de ne and imple-ment hybrid systemsrdquo (p 1) None of this is served up for statisticians butthere are some useful introductions Chapter 2 provides an introduction to theentire body of methods called ldquofuzzy logicrdquo Chapter 4 gives an overview ofneural networks multilayer percepterons nonsupervised networks and radialbasic functions Chapter 6 similarly presents evolutionary-type optimizationalgorithms concentrating mainly on genetic annealing Subsequent chaptersextend these basic concepts to chaotic systems neuro-fuzzy networks andfuzzy systems optimization by combining the fundamental methodologies toproduce more powerful technologies

Some statisticians are using these technologies in their research into cre-ating new statistical tools More importantly nonstatisticians are using thesetools instead of more traditional statistical tools For example the chemicaland pharmaceutica l literature contains a number of articles on how geneticalgorithms can provide a better process for optimization through experimen-tation than conventiona l experimental design (see eg Webster-Botz 2000)The book provides an opportunity for statisticians to get a good introductionto powerful computing tools that may help them solve their own problems

REFERENCES

Chiang L Russell E and Braatz R (2001) Fault Detection and Diagnosisin Industrial Systems New York Springer-Verlag

Chonavel T and Vaton S (2001) Statistical Signal Processing New YorkSpringer-Verlag

Webster-Botz D (2000) ldquoExperimental Design for Fermentation MediaDevelopment Statistical Design or Global Random Searchrdquo Journal ofBioscience and Bioengineering 90 473ndash483

Making Hard Decisions (2nd ed) by Robert T Clemen

and Terence Reilly Paci c Grove CA Duxbury 2001ISBN 0-534-36597-3 xxv C 733 pp C CD $9495

The rst edition of this book was reviewed for Technometrics by Liu (1992)An earlier version of this second edition was published in 1996 It was not allthat different from the rst edition so I never got enough interest in a reporton a new edition actually to do it Concerning the rst edition Liu concludedldquo I highly recommend this book as a textbook for the rst course in decisionanalysis provided that the instructor and students are also equipped with andfamiliar with some of the personal computer decision software that Clemenused to illustrate his examples and problemsrdquo

This book is neither a new book nor a new edition Perhaps nally respond-ing to Liursquos stipulation concerning software the authors note that ldquothis ver-sion focuses on the use of an electronic spreadsheet as a platform for modelingand analysisrdquo (p xxii) Obviously any business student in the 21st century


200 BOOK REVIEWS

will be familiar with Excel The book is subtitled ldquoWith Decision Toolsiexcl rdquowhich describes the selected spreadsheet software a set of ve programsmarketed by the Palisade Corporation (wwwpalisadecom) These include thewidely used RISK program for risk analysis using Monte Carlo simulation

The second author apparently undertook the responsibility for interfacingthe Decision Tools suite with the existing content of the book The introduc-tory chapter explains where the software ts in Many of the chapters concludewith sections on using the software to do the operations in the book that needsoftware These sections provide the students with the software familiaritymentioned in the conclusion of Liursquos review

Liu (1992) gave a very complete summary of the content of the rst edi-tion which generally still applies to the content of this version of the secondedition Liursquos review is also interesting for the perspective that he brings torisk analysis The review does not bring all the bookrsquos details to the surfaceand does not convey the richness of the case studies and the illustrationstherein This emphasis on case studies is the perspective that one of my BPcolleagues provides in his recently published book (Koller 2000) a virtuallitany of the vast spectrum of applications that can bene t from the use of riskanalysis

REFERENCES

Koller G (2000) Risk Modeling for Determining Value and Decision MakingBoca Raton FL Chapman and HallCRC

Liu S (1992) Review of Making Hard Decisions by R Clemen Techno-metrics 34 365ndash366

Applying Statistics in the Courtroom by Phillip I GoodBoca Raton FL Chapman and HallCRC 2001 ISBN1-58488-271-9 xviii C 276 pp $6995

There has been a steady ow of books about statistics for jurisprudence Most recently in these reports was a discussion of the book by Finkelsteinand Levin (2001) the rst of these legal statistics books to appear as a secondedition My father was a trial judge my company is always involved in sometype of litigation and my cousin manages data analysis for a successful expertwitness practice so knowledge of this literature seems to me essential for anyindustrial statistician This is the best book on statistics and the law that hasyet been written

The author has written this text ldquofor two audiences with a single commongoal to ensure attorneys and statisticians will work together successfully onthe application of statistics in the lawrdquo (p v) Subtitled ldquoA New Approachfor Attorneys and Expert Witnessesrdquo the book presents ldquothe principles ofstatistics and probability not as a series of symbols but in the words of thejuristsrdquo (ibid) There are only a handful of equations in the entire text Thisauthor has written a number of advanced statistics books most recently Good(2000) reported in Ziegel (2001) and Good (1999) reviewed in Chaubey(2000) so the approach does not re ect any lack of statistics capability onthe part of the author

The book comprises four parts Part I ldquoSamples and Populationsrdquo containsthree chapters about sampling methods and a fourth chapter about descrip-tive statistics Some of this material deals with jury selection other materialconcerns the collection and analysis of data Part II has four chapters on prob-ability Topics include logic independence conditional probabilities and evenBayesrsquos theorem Three applications chapters discuss litigation from criminallaw civil law and environmenta l hazards The latter chapter includes riskanalysis

Part III ldquoHypothesis Testing and Estimationrdquo contains four chapters aboutthe statistical analysis of data The rst chapter deals with statistical signi -cance The second chapter includes distribution theory permutation tests andcontingency tables among its topics The last two chapters cover correlationand regression Part IV ldquoApplying Statistics in the Courtroomrdquo concludesthe book with three interesting chapters The rst of these deals with thepractice of ldquopreventive statisticsrdquo including effective use of controls repre-sentative random samples or adequate sample sizes by lawyers and statisticalscientists The next chapter ldquoWhat Every Statistician Should Know AboutCourtroom Proceduresrdquo and the last chapter ldquoMaking Effective Use of Statis-tics and Statisticiansrdquo provide material that should be fully understood byevery industrial statistician and their lawyer colleagues Unless your company

never appears in any adversarial proceeding this book should be on yourbookshelf It is profusely illustrated with a wealth of cases that will interestany statistician

REFERENCES

Chaubey Y (2000) Review of Resampling Methods A Practical Guide toData Analysis by P Good Technometrics 42 311

Finkelstein M and Levin B (2001) Statistics for Lawyers (2nd ed)New York Springer-Verlag

Good P (1999) Resampling Methods A Practical Guide to Data AnalysisBoston Birkhaumluser

(2000) Permutation Tests (2nd ed) New York Springer-VerlagZiegel E (2001) Editorrsquos Report for Permutation Tests (2nd ed) by P Good

Technometrics 43 114

Statistics for Lawyers (2nd ed) by Michael O Finkel-

ste in and Bruce Levin New York Springer-Verlag 2001ISBN 0-587-95007-9 xxx C 617 pp $7995

Statistics for legal applications is a new genre that is becoming increasinglymore commonplace This is the rst book on statistics for lawyers that hascome out in a second edition Statistics books for lawyers have been gettingever better In this issue I give glowing praise to the ne new book by Good(2001) When it appeared the rst edition of this book was my pick as thebest available book of its kind It was also the rst book to fully integratestatistical methodology with its use in litigation in the courts In the reviewof the rst edition (Ziegel 1992) I wrote ldquo I strongly recommend this bookfor any statisticians who must help lawyers understand statistics or who areconsulting with lawyers to support their efforts with statistical assistance Thisbook has been carefully prepared to provide cogent and succinct informationabout a broad spectrum of statistical methods applicable to jurisprudence There is complete linkage to actual cases in which statistics has played a rolein the judicial processrdquo My description of the rst edition was fairly completeby my editorial report standards

The most impressive change in the book can be attributed to the pub-lisher who has included the book in its relatively new series on Statisticsfor Science and Public Policy First published as an inexpensive softcoveredition the book is now a beautiful hardcover volume at twice the price ofthe rst edition although it is certainly not overpriced The most signi cantimprovement is the layout The statistical content for each of the chapterscreates the sections and the cases provide subsections In the rst editioneverything was a section The authors highlight the new topics for the cases inthe second edition ldquoDNA evidence epidemiologic studies in toxic substancelitigation statistical models for adjusting census counts and vote dilutioncasesrdquo (p vii) There is an entirely new chapter on statistics in epidemiologyOther new content includes Monte Carlo methods additional probability dis-tributions (geometric and exponential) and several new regression methodsThe very long regression chapter was split and the second chapter on themore complex methods now covers time series models and locally weightedregression

More comprehensive than the book by Good (2001) and authored jointlyby a lawyer and a statistician this book continues to be a wonderful resourcefor the use of statistics in the courts It would be wonderful to know whatmy father now deceased a trial judge for 36 years in county and state courtsin Ohio would think of this book and his need for some comprehensionof statistical methods in understanding testimony and rendering decisionsMaybe he would even be more impressed now with his son who did notchoose to become a lawyer

REFERENCES

Good P (2001) Applying Statistics in the Courtroom Boca Raton FL Chap-man and HallCRC

Ziegel E (1992) Editorrsquos Report for Statistics for Lawyers by M Finkelsteinand B Levin Technometrics 34 122ndash123

(2002) Editorrsquos Report for Applying Statistics in the Courtroom byP Good Technometrics 44 forthcoming


BOOK REVIEWS 201

Statistics for Environmental Science and Managementby Bryan F J Manly Boca Raton FL Chapman andHallCRC 2001 ISBN 1-58488-029-5 x C 326 pp$4995

As manager for statistics technology support across a giant corporation Ihave a number of distinct clienteles One is our cadre of environmental scien-tists who work both in centralized groups and embedded within businessesThese people frequently have statistics applications usually have a some basicstatistics class in college and generally are totally unprepared for the uniqueaspect of the statistical data in environmental science Manly wrote this bookfor these people

A former faculty member who has written books for environmenta l sciencemost recently Manly (1997) reported in Ziegel (1998) and now a statisticianfor an environmenta l consulting company the author knows the needs of envi-ronmental scientists He also knows how to effectively present methodologyto his audience Presuming some previous basic statistics course the authorjumps right into sampling after the usual introductory chapter for motivationby examples Focused more on traditional sampling than the excellent bookby Thompson (1992) the chapter does not include sampling for spatial dataNext is a chapter on various types of statistical models for data discretedistributions continuous distributions linear regression analysis of variance(ANOVA) and generalized linear models

The subsequent four chapters approach statistical methodology from theperspective of the environmenta l scientist The rst of these ldquoDrawing Con-clusions from Datardquo discusses types of studies types of inference signi -cance tests con dence intervals randomization tests bootstrapping multipletesting meta-analysis and Bayesian inference all without much illustrationNext comes ldquoEnvironmental Monitoringrdquo which begins with some examplesof spatial designs Methods presented for detecting change include ANOVAShewhart control charts and cumulative sum (CUSUM) charts This chapteris mostly illustrations The next chapter ldquo Impact Assessmentrdquo focuses onbefore-after impact-control and impact-gradient designs The last chapter inthis group ldquoAssessing Site Reclamationrdquo has bioequivalence as its primarytopic Here bioequivalence refers to biological equivalence after a site isimpacted

The bookrsquos last four chapters present special statistical tools for environ-mental data that lie outside the scope of an introductory statistics course Firstis ldquoTime Series Analysisrdquo which focuses primarily on serial correlation versusrandomness and the detection of change points and trends Additional topicsinclude autoregressive integrated moving average models frequency domainanalysis and forecasting More material is needed on seasonality Next comesldquoSpatial Data Analysisrdquo which focuses mostly on quadrat counts and pointpatterns but concludes with several sections on geostatistics ldquoCensored Datardquocovers the important issue of nondetection in chemicals measurement Last isa chapter on ldquoMonte Carlo Risk Assessmentrdquo

Overall this book has a huge variety of topics and numerous examples inmost of the chapters There are excellent references throughout There is somediscussion and even illustration of statistical computing Hopefully the secondedition will incorporate more computing but readers can use the book byMillard (2001) reviewed by Lumley (2001) as a guide This book is a greatand inexpensive library addition for statisticians and environmental scientistswho analyze environmenta l data

REFERENCES

Lumley T (2001) Review of Environmental Statistics With S-PLUS byS Millard and N Neerchal Technometrics 43 495

Manly B (1997) Randomization Bootstrap and Monte Carlo Methods inBiology (2nd ed) Boca Raton FL Chapman and Hall

Millard S and Neerchal N (2001) Environmental Statistics With S-PLUSBoca Raton FL CRC

Thompson S (1992) Sampling New York WileyZiegel E (1998) Editorrsquos Report for Randomization Bootstrap and Monte

Carlo Methods in Biology (2nd ed) by B Manly Technometrics 40 84

Bayesian Survival Analysis by Joseph G Ibrahim Ming-Hui Chen and Debajyoti Sinha New York Springer-Verlag 2001 ISBN 0-387-95277-2 xiv C 479 pp $7995

As recent reports and reviews have noted books on Bayesian statisticsare now routinely appearing in second editions such as that by Carlin

and Lewis (2000) reported by Ziegel (2001) Books on survival analysisappear regularly examples include those by Therneau and Grambsch (2000)reported by Lin and Zelterman (2002) or Hougaard (2000) reviewed byKenyon (2002) As the latter book and this book demonstrate survivalanalysis can be expanded beyond a basic set of models Those books likethis one are driven by applications ldquoessentially from the health sciencesincluding cancer AIDS and the environmentrdquo (p viii) The author beginsthe book by noting that ldquothe analysis of time-to-event data generally calledsurvival analysis arises in many elds of etuding includingcent cent cent engineeringrdquoFortunately we do not have many instances of failure events for themechanical equipment in the petrochemical industries but that situation alsolimits the use of modeling Unfortunately there are many more instancesof multiple myeloma melanoma breast cancer and the other diseases thatmotivate the book in its rst chapter

This is not an elementary book The back cover copy mentions ldquosecond-or third-year graduate studentsrdquo It is not a theorem-and-proof presentationThree of the ten chapters have an appendix into which the serious proofs havebeen dumped The book develops methodology and does this at a high levelbecause the reader is presumed to have a mathematical statistics backgroundin both classical and Bayesian methods Happily the book is replete withexamples This is one of the best combinations of advanced methodology andpractical applications that I have encountered

The introductory chapter gives motivation for sampling posterior distri-butions and determining appropriate prior distributions The authors arguefor the advantages of Bayesian inference Particularly effective argumentsare the ability to handle estimation for messy censored data with MarkovChain Monte Carlo sampling and of course using the prior information thatwould always exist for the equipmental in a chemical process unit The nextfour chapters each present a different model category parametric models(ie exponentia l Weibull extreme value lognormal piecewise gamma pro-cess beta process Cox Dirichlet) semiparametric models frailty models (forunknown or unobservable risk factors) and cure rate models (for situationswhere some incidences eventually become cures) The latter two chapters bothlead to the use of multivariate models The next three chapters are extensionscomparison of models joint models with survival and longitudinal data andmodels for missing covariate data Chapter 9 explains the Bayesian approachto the design and monitoring of randomized clinical trials The nal chapterhas a long list of advanced topics including accelerated failure time modelsand building models with MARS and neural networks It includes 11 differentexamples Computing support for the book comes from the package calledBUGS although the book never tells how to set up BUGS to produce theresults

REFERENCES

Carlin B and Louis T (2000) Bayes and Empirical Bayes Methods forData Analysis (2nd ed) Boca Raton FL Chapman and HallCRC

Hougaard P (2000) Analysis of Multivariate Survival Data New YorkSpringer-Verlag

Kenyon J (2002) Review of Analysis of Multivariate Survival Data byP Hougaard Technometrics 44 86ndash87

Lin H and Zelterman D (2002) Review of Modeling Survival Data byT Therneau and P Grambsch Technometrics 44 85ndash86

Robert C (2001) The Bayesian Choice (2nd ed) New York Springer-Verlag

Therneau T and Grambsch P (2000) Modeling Survival Data New YorkSpringer-Verlag

Ziegel E (2001) Editorrsquos Report on Bayes and Empirical Bayes MethodsFor Data Analysis (2nd ed) by B Carlin and T Louis Technometrics43 246

Analyzing Medical Data Using S-PLUS by B SEveritt and Sophia Rabe-Hesketh New York Springer-Verlag 2001 ISBN 0-387-98862-9 xii C 485 pp $7995

Books on medicine have been reported here often Books on S-PLUS havebeen reported less frequently This book is being presented as an appliedstatistics book that is highly recommended to any statistician who might bean S-PLUS beginner like me I recently took a class on S-PLUS for SASusers The class was nothing special but I did develop an enthusiasm forusing S-PLUS


202 BOOK REVIEWS

This book begins with an overview of S-PLUS This is completely directedtoward the command language and would do nothing to convince a nonuserthat taking up S-PLUS would be an enjoyable task The rest of the book iswhat is special for the S-PLUS novice There are 19 chapters devoted to awide variety of methods for the statistical analysis of data These chapters donot present any theory or even develop any methodology Rather the authorsexplain what the analysis is all about give some examples of when it wouldbe used show an outline of the methodology in a couple of pages of ldquodisplayrdquoboxes present the results of using S-PLUS for the methods to analyze thedata from the examples provide discussion of the results as their text and listscript les for all of the S-PLUS commands at the end of each chapter Aswith doing graphics in SAS using the command language in S-PLUS seemsto be easiest when modifying existing scripts

The rst author is an exceptionally proli c textbook writer who teachesat an institute for psychiatry [see eg Everitt and Dunn (2001) reported inZiegel (2002a) and Everitt and Pickles (1999) reported in Ziegel (2002b)]In addition to covering conventiona l and traditional methodologies such asgraphics comparison tests contingency tables regression and linear modelsthe book has chapters devoted to smoothing generalized linear models mixedmodels generalized additive models nonlinear models regression trees andsurvival analysis Multivariate analysis also is included with chapters onprincipal components and factor analysis cluster analysis and discriminantanalysis There is an appendix the S-PLUS (GUI) The GUI has been consid-erably enhanced in version 6 However the GUI generally seems to be viewedwith disdain by experienced S-PLUS users Although only a few pages areprovided here the GUI does not require much instruction

If I were ever to write a book for statistics in industry this would be a goodmodel I would certainly write for an audience that would use only a GUIinterface I am not big on the need for programming expertise by statisticssoftware users I question the necessity for medical researchers to know howto write programs in S-PLUS Nonetheless my perspective does not detractfrom my good opinion of a well-organized nicely conceived book that willbe useful to practitioners with S-PLUS in any environment

REFERENCES

Everitt B and Dunn G (2001) Applied Multivariate Data Analysis (2nded) New York Oxford University Press

Everitt B and Pickles A (1999) Statistical Aspects of Design and Analysisof Clinical Trials London Imperial College Press

Ziegel E (2002a) Editorrsquos Report on Statistical Aspects of Design andAnalysis of Clinical Trials by B Everitt and A Pickles Technometrics44 forthcoming

Ziegel E (2002b) Editorrsquos Report on Applied Multivariate Data Analysis(2nd ed) by B Everitt and G Dunn Technometrics 44 forthcoming

Biostatistics in Clinical Trials edited by Carol Redmond

and Theodore Colton Chichester UK Wiley 2001ISBN 0-471-82211-6 xx C 501 pp $19500

This is the second volume in the publisherrsquos Reference Series in Biostatis-tics The rst volume by Gail and Benichou (2000) is reported on elsewherein this volume (Ziegel 2002) In these new volumes different editors havemade extracts from the publisherrsquos Encyclopedia of Biostatistics (Armitageand Colton 1998) a six-volume opus reported for Technometrics by Ziegel(2000) The extract volumes are focused on a speci c subject In additioneach of the extracted articles has been updated and some new articles havebeen added to provide the completeness necessary for a single-subject ency-clopedia

Like the predecessor volume on epidemiology this is just another ency-clopedia not withstanding the missing appellation in the title At 490 pagesthis volume is only half the size of the epidemiology volume This ought toreduce the price in half but at $295 this volume costs considerably more thanhalf of the $395 price of the rst volume This volume has 17 new entriesor essays nearly twice as many new entries as the epidemiology volume Myfavorite entry is the annotated bibliography of most of the existing textbooksfor clinical trials Other new essays speci c to this particular volume on clin-ical trials include entries for Bayesian methods bene trisk assessment andsoftware The rst two of those add 12 pages at the beginning of the Brsquos

In 2000 I reviewed the content of some typical entries for the encyclopedi a(Ziegel 2000) The editors of this volume state in the Preface that ldquowe decidednot to include articles on basics nor articles on methods and analysis thatwe viewed as not particular to clinical trials although commonly employedin the analysis of clinical trial resultsrdquo There is after all a 10-page discussionon textbooks Elsewhere in this volume I review the book on clinical trials byEveritt and Pickles (2000)

REFERENCES

Armitage P and Colton T (eds) (1998) Encyclopedia of BiostatisticsVols 1ndash6 New York Wiley

Everitt B and Pickles A (2000) Statistical Aspects of the Design andAnalysis of Clinical Trials River Edge NJ World Scienti c

Gail M and Benichou J (2000) Encyclopedia of Epidemiologic MethodsWest Sussex UK Wiley

Ziegel E (2000) Editorrsquos Report on Encyclopedia of Biostatistics Vols 1ndash6edited by P Armitage and T Colton Technometrics 42 222

(2002a) Editorrsquos Report on Statistical Aspects of the Design andAnalysis of Clinical Trials by B Everitt and A Pickles Technometrics44 forthcoming

(2002b) Editorrsquos Report on Encyclopedia of Epidemiologic Methodsby M Gail and J Benichou Technometrics 44 forthcoming

Encyclopedia of Epidemiologic Methods by MitchellH Gail and Jacques Benichou West Sussex UKWiley 2000 ISBN 0-471-86641-5 xxi C 978 pp $395

This book is an excellent effort by the publisher to generate some addi-tional income from one of its most valuable properties the Encyclopedia ofBiostatistics (Armitage and Colton 1998) Encompassing six volumes 5000pages and 1200 articles that work was reported on by Ziegel (2000) whonoted that the price currently $3775 would put it well beyond the reach of allindividuals and probably even most libraries Here the publisher has launchedthe Wiley Reference Series in Biostatistics which has the two editors for theEncyclopedia of Biostatistics as its series editors As the authors state ldquoAlarge majority of the articles have been extracted from John Wileyrsquos Encyclo-pedia of Biostatistics with some updating of references and coverage of morerecent developmentsrdquo (p vii) There also are some additional articles addedparticularly for their explicit focus on epidemiology such as the epidemiologyoverview article

My earlier report on the Encyclopedia of Biostatistics (Ziegel 2000)describes the content of the individual articles in some detail and generallylavishes high praise on the whole undertaking Extracting some of thematerial for a speci c subject area is a fantastic idea Unfortunately theeditors of this epidemiology volume were very inclusive in their selectionsAlmost 1000 of the 5000 pages from the encyclopedia have landed in thisvolume making it a very large book It also is very costly ($395) probablystill beyond the reach of most individuals and maybe even libraries Thatsaid though the price is very reasonable for nearly 20 of the content of anencyclopedia that costs $3775

Along with the bookrsquos epidemiology overview I liked the 18-page articleon the determination of sample sizes for epidemiology studies and the newentry for software in epidemiology There is an entertaining new essay onthe ldquoSex Ratio at Birthrdquo I am happy to have a condensed version of theEncyclopedia for an area relevant to the petrochemical industry It would benice if somehow a few readers could justify buying a copy for their libraryso perhaps more modest extractions should be made

REFERENCES


Ziegel E (2000) Editorrsquos Report on Encyclopedia of Biostatistics Vols 1ndash6by P Armitage and T Colton Technometrics 42 222ndash223


BOOK REVIEWS 203

Forthcoming Reviews

Books listed here have been assigned for review in the past quarterPublication of their reviews or reports generally would occur within thenext four issues of the journal Persons interested in reviewing speci cbooks must notify the editor by the publication date for the book Personsinterested in being reviewers should contact the editor by electronic mail(ziegelerbpcom)

Analysis of Messy Data Volume III Analysis of Covariance by George AMilliken and Dallas E Johnson Chapman amp HallCRC

Box on Quality and Discovery edited by George C Tiao Soslashren BisgaardWilliam J Hill Daniel Pentildea and Stephen M Stigler John Wiley

Calculated Bets by Steven Skiena Cambridge University Press

Computational Statistics Handbook with MATLABacirc by Wendy L Martinezand Angel R Martinez Chapman amp HallCRC

A Contingency Table Approach to Nonparametri c Testing by J C W Raynerand D J Best Chapman amp HallCRC

Data Mining and Uncertain Reasoning by Zhengxin Chen John Wiley

The Elements of Statistics by James B Ramsey Duxbury Press

Eliciting and Analyzing Expert Judgment by Mary A Meyer and Jane MBooker ASA-SIAM

Experimental Design with Applications in Management Engineering and theSciences by Paul D Berger and Robert E Maurer Duxbury Press

Foundations of Time Series Analysis and Prediction Theory by MohsenPourahmadi John Wiley

Frontiers in Statistical Quality Control 6 edited by H-J Lenz and P-ThWilrich Physica-Verlag

Generalized Linear Models by Raymond H Myers Douglas C Montgomeryand G Geoffrey Vining John Wiley

A Handbook of Statistical Analyses using SAS (2nd ed) by Geoff Der andBrian S Everitt Chapman amp HallCRC

In All Likelihood by Yudi Pawitan Oxford University Press

Interpreting ISO 90002000 with Statistical Methodology by James L Lam-precht ASQ Quality Press

An Introduction to Generalized Linear Models (2nd ed) by Annette J Dob-son Chapman amp HallCRC

Multilevel Modelling of Health Statistics edited by A H Leyland andH Goldstein Wiley

Practical Experiment Designs (3rd ed) by William J Diamond John Wiley

Principles of Medical Statistics by Alvan R Feinstein Chapman amp HallCRC

Probability Statistical Optics and Data Testing (3rd ed) by B R FriedenSpringer-Verlag

Probability with Statistical Applications by Rinaldo B Schinazi BirkhaumluserSpringer-Verlag

Resampling Methods (2nd ed) by Phillip I Good BirkhaumluserSpringer-Verlag

Sampling Methodologies with Applications by Poduri S R S Rao Chapmanamp HallCRC

Should We Risk It by Daniel M Kammen and David M Hassenzahl Prince-ton University Press

Single-Case and Small-n Experimental Designs by John B Todman and PatDugard Lawrence Erlbaum

Six Sigma and Beyond Foundations of Excellent Performance by D HStamatis CRC Press

Six Sigma and Beyond Problem Solving and Basic Mathematics by D HStamatis CRC Press

Statistical Methods for Clinical Trials by Mark X Norleans Marcel Dekker

Statistics in Plain English by Timothy C Urdan Lawrence Erlbaum

Statistics in the 21st Century edited by Adrian E Raftery Martin A Tannerand Martin T Wells Chapman amp HallCRC

Survey Nonresponse edited by Robert M Groves Don A Dillman John LEltinge and Roderick J A Little Wiley

The Visual Display of Quantitative Information (2nd ed) by Edward R TufteGraphics Press

Weighing the Odds by David Williams Cambridge University Press


186 BOOK REVIEWS

Analyzing Medical Data Using S-PLUSB S Everitt and Sophia Rabe-Hesketh 201

Biostatistics in Clinical TrialsCarol Redmond and Theodore Colton (editors) 202

Encyclopedia of Epidemiologic Methods

Mitchell H Gail and Jacques Benichou 202

Forthcoming Reviews 203

Modern Statistics for Engineering and Quality Improve-ment by John Lawson and John Erjavec Paci c GroveCA Duxbury 2001 ISBN 0-534-19050-2 xi C 810 pp$7195

The authors set out to produce an undergraduat e text and reference thatfocuses on what they feel are the most useful statistical tools encounteredby practicing engineers They succeeded in creating a valuable book rich inexperimental design and response surface methods (458 pages 10 chapters)augmented by standard coverage of quality control (134 pages) but with onlyminimal treatment of basic probability and statistics (132 pages) This is farmore a book on experimental design application in industry than it is a gen-eral presentation of modern statistical methods for engineers as found in forexample the books by Montgomery and Runger (1999) and Ostle TurnerHicks and McElrath (1996)

The bookrsquos 5 parts and 18 chapters are as follows

Part I Introduction1 The Scienti c Method and Statistics2 Concepts of Quality Control

Part II Basic Tools3 Theoretical Background Probability and Statistics4 Descriptive Tools5 Probability Plots6 Inferential Statistics Prediction and Statistical Decision Rules

Part III Good Experiments Make for Good Statistics7 Strategies for Experimentation with Multiple Factors8 Basic Two-Level Factorial Experiments9 Additional Tools for Design and Analysis of Two-Level Factorials

10 Regression Analysis11 Multiple Level Factorial Experiments12 Screening Designs

Part IV Optimization Experiments13 Response Surface Methodology14 Response Surface Model Fitting15 Mixture Experiments

Part V Variability and Quality16 Characterizing Variability in Data17 Shewhart Control Charts18 Off-Line Quality Control and Robust Design

The book begins with an engaging introduction supported by several exam-ples that should alert engineering students that valuable information is con-tained within The book immediately challenges students to participate inengineering thought using Pareto process ow and cause-and-effec t diagramsto address defects in circuit boards Quality problems in transmission castingsand the copperplating of ceramic substrates further develop the need for under-standing the physical situation through reasoned data collection and analysisAn erroneous gure (Fig 27) and table reference (Table 23) partially con-fuses the message of the rst motivating example in Chapter 2 involvingcontrol charting and tool wear

Part II expects much of students Discrete random variables and distribu-tions appear only ve pages into Chapter 3 with the binomial distributionjust three pages later In all only 36 pages are devoted to an introduction toprobability random variables and their distributions (including the binomial

Poisson uniform exponential normal and lognormal) and the law of largenumbers and central limit theorem This streamlined presentation requires anoccasional leap for the student For example the expected value of a linearcombination of independent variables Y D aY1 C bY2 is af rmed beginningwith

E4Y 5 D E4aY1 C bY25 DZZ

4ay1 Cby25f14y15f24y25 dy1 dy20

But the only previous mention of joint probability let alone the relationshipbetween joint and marginal densities under independence comes from themultiplication rule P4AampB5 D P4A5P4B5 An equally brief but less intimi-dating presentation of summary statistics and usual graphical characterizationsof data follows in Chapter 4 Chapter 5 presents a very readable and generaltreatment of probability plots and their use and Chapter 6 gives a clear butconcise introduction

Parts III and IV are the real strength of this book supported by ampletables of design schemes in the appendixes Topics ow naturally in answer-ing the question of what needs to be done or considered next to pursue thequality engineering problem at hand The many examples are splendid Thechapter titles specify most topics but to suggest the thoroughnes s of cover-age speci c topics include PlacketndashBurman designs mixed-level fractionalfactorial designs sequential experimentation matrix algebra for regressionBoxndashCox transformations split-plot designs central composite designs andBoxndashBehnken designs

Part V concludes with more experimental designs involving variance com-ponents measurement error and staggered nested designs in Chapter 16 TwoChapters 17 and 18 continue the discussion of control charts begun in PartI and extend it to off-line quality control and robust design Chapter 18 alsoincludes parameter design experiments

In a one-semester format this book would be appropriate for upper-levelengineering students who have had some previous exposure to basic statis-tics The authors concede that Part II is streamlined to provide more timefor exploring topics of greater interest but they assert that greater detail inthat section is not required to successfully move to the latter sections of thebook Perhaps so but the instructor should be prepared to augment if neededThe authors suggest course outlines of selected chapters dependent on theengineering interest but I think a book as comprehensive as this would bebest served over two semesters As a reference for practicing engineers itworks well

Barry A Bodt

US Army

REFERENCES


Ostle B Turner K Hicks C and McElrath G (1996) Engineering Statis-tics The Industrial Experience Belmont CA Wadsworth

Multivariate Analysis of Quality An Introduction byHarald Martens and Magni Martens West Sussex UKWiley 2001 ISBN 0-471-97428-5 xx C 445 pp $135

The authors state that this book has been written to meet a need for anentry-level textbook on multivariate data analysis for the practical researcherin academia or industry The text is presented as a narrative analysis of mul-tivariate data using soft modeling The authors have little use for classicalstatistical analysis stating that (1) ldquothe book makes little or no use of tra-ditional statistical distribution theory and takes a dim view on hypothesistestingrdquo (2) ldquotraditional hypothesis testing methods in mathematical statisticsfocus more on the noise than on the interesting signal in datardquo and (3) ldquocom-puterised statistical methods for re-sampling have already reduced the needfor teaching complicated mathematical statistics to unwilling and terri ednon-statisticiansrdquo

So how do these chemometricians predict a response matrix Y from amatrix of x-predictor variables X First matrix X is represented by the prod-uct of a score matrix TA (ie ldquoArdquo principal components ) and a loading matrixP plus an error matrix Second matrix Y is modeled by the product of theTA score matrix and a loading matrix Q plus an error matrix The Y matrix isalso represented by the linear regression of X with coef cient matrix B plus


BOOK REVIEWS 187





Charles K Bayne










188 BOOK REVIEWS



Phillip Yates











Lora Zimmer







BOOK REVIEWS 189









Timothy Robinson


REFERENCES










Melvin Alexander

Qualistics


190 BOOK REVIEWS

REFERENCES
























Margaret A Nemeth

Monsanto Company








Margaret A Nemeth

Monsanto Company


BOOK REVIEWS 191





Subir Ghosh


REFERENCES


















192 BOOK REVIEWS








J Brian Gray


REFERENCES





















BOOK REVIEWS 193







Henry W Altland

Corning Inc

REFERENCE

















194 BOOK REVIEWS










J Charles Kerkering










Thomas E Wehrly


REFERENCE





BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































BOOK REVIEWS 187





Charles K Bayne










188 BOOK REVIEWS



Phillip Yates











Lora Zimmer







BOOK REVIEWS 189









Timothy Robinson


REFERENCES










Melvin Alexander

Qualistics


190 BOOK REVIEWS

REFERENCES
























Margaret A Nemeth

Monsanto Company








Margaret A Nemeth

Monsanto Company


BOOK REVIEWS 191





Subir Ghosh


REFERENCES


















192 BOOK REVIEWS








J Brian Gray


REFERENCES





















BOOK REVIEWS 193







Henry W Altland

Corning Inc

REFERENCE

















194 BOOK REVIEWS










J Charles Kerkering










Thomas E Wehrly


REFERENCE





BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































188 BOOK REVIEWS



Phillip Yates











Lora Zimmer







BOOK REVIEWS 189









Timothy Robinson


REFERENCES










Melvin Alexander

Qualistics


190 BOOK REVIEWS

REFERENCES
























Margaret A Nemeth

Monsanto Company








Margaret A Nemeth

Monsanto Company


BOOK REVIEWS 191





Subir Ghosh


REFERENCES


















192 BOOK REVIEWS








J Brian Gray


REFERENCES





















BOOK REVIEWS 193







Henry W Altland

Corning Inc

REFERENCE

















194 BOOK REVIEWS










J Charles Kerkering










Thomas E Wehrly


REFERENCE





BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































BOOK REVIEWS 189









Timothy Robinson


REFERENCES










Melvin Alexander

Qualistics


190 BOOK REVIEWS

REFERENCES
























Margaret A Nemeth

Monsanto Company








Margaret A Nemeth

Monsanto Company


BOOK REVIEWS 191





Subir Ghosh


REFERENCES


















192 BOOK REVIEWS








J Brian Gray


REFERENCES





















BOOK REVIEWS 193







Henry W Altland

Corning Inc

REFERENCE

















194 BOOK REVIEWS










J Charles Kerkering










Thomas E Wehrly


REFERENCE





BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































190 BOOK REVIEWS

REFERENCES
























Margaret A Nemeth

Monsanto Company








Margaret A Nemeth

Monsanto Company


BOOK REVIEWS 191





Subir Ghosh


REFERENCES


















192 BOOK REVIEWS








J Brian Gray


REFERENCES





















BOOK REVIEWS 193







Henry W Altland

Corning Inc

REFERENCE

















194 BOOK REVIEWS










J Charles Kerkering










Thomas E Wehrly


REFERENCE





BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































BOOK REVIEWS 191





Subir Ghosh


REFERENCES


















192 BOOK REVIEWS








J Brian Gray


REFERENCES





















BOOK REVIEWS 193







Henry W Altland

Corning Inc

REFERENCE

















194 BOOK REVIEWS










J Charles Kerkering










Thomas E Wehrly


REFERENCE





BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































192 BOOK REVIEWS








J Brian Gray


REFERENCES





















BOOK REVIEWS 193







Henry W Altland

Corning Inc

REFERENCE

















194 BOOK REVIEWS










J Charles Kerkering










Thomas E Wehrly


REFERENCE





BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































BOOK REVIEWS 193







Henry W Altland

Corning Inc

REFERENCE

















194 BOOK REVIEWS










J Charles Kerkering










Thomas E Wehrly


REFERENCE





BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































194 BOOK REVIEWS










J Charles Kerkering










Thomas E Wehrly


REFERENCE





BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































BOOK REVIEWS 195













Craig B Borkowf










196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































196 BOOK REVIEWS



Pradipta Sarkar









Maliha S Nash








BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































BOOK REVIEWS 197




Yachen Lin







Graham J Wills


REFERENCE




Eric R Ziegel

BP






REFERENCES








198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































198 BOOK REVIEWS



REFERENCES












REFERENCES









BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































BOOK REVIEWS 199

REFERENCES









REFERENCES





York McGraw-Hill







REFERENCES









200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































200 BOOK REVIEWS




REFERENCES









REFERENCES











REFERENCES





BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































BOOK REVIEWS 201







REFERENCES











REFERENCES











202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































202 BOOK REVIEWS




REFERENCES










REFERENCES











REFERENCES




BOOK REVIEWS 203

Forthcoming Reviews




































BOOK REVIEWS 203

Forthcoming Reviews




































Book Reviews - JISCMail

Documents