Short Book Reviews On-Line 2004

THE ANALYSIS OF TIME SERIES. AN INTRODUCTION, 6th edition.	C. Chatfield.
NONLINEAR TIME SERIES: NONPARAMETRIC AND PARAMETRIC METHODS.	J. Fan and Q. Yao.
CASE STUDIES IN RELIABILITY AND MAINTENANCE.	W.R. Blischke and D.N. Prabhakar Murthy (Eds.).
EXPLORATORY DATA MINING AND DATA CLEANING.	T. Dasu and T. Johnson.
STATISTICAL MODELING AND ANALYSlS FOR DATABASE MARKETING: EFFECTIVE TECHNIQUES FOR MINING BIG DATA.	B. Ratner.
APPLIED DATA MINING: STATISTICAL METHODS FOR BUSINESS AND INDUSTRY.	P. Giudici.
AN INTRODUCTION TO CREDIT RISK MODELING.	C. Bluhm, L. Overbeck and C. Wagner.
FOUNDATIONS OF RISK ANALYSIS. A Knowledge and Decision-Oriented Perspective.	T. Aven.
RISK ANALYSIS IN ENGINEERING AND ECONOMICS.	B.M. Ayyub.
RISK ANALYSIS IN FINANCE AND INSURANCE.	A. Melnikov, Translated and edited by A. Filinkov.
WEAK CONVERGENCE OF FINANCIAL MARKETS.	J.L. Prigent.
INFINITE DIVISIBILITY OF PROBABILITY DISTRIBUTIONS ON THE REAL LINE.	F.W. Steutel and K. van Harn.
MARKOV PROCESSES FROM K. ITÔ'S PERSPECTIVE.	D.W. Stroock.
TAKING CHANCES: WINNING WITH PROBABILITY, New Edition. Including Who Wants to be a Millionaire and the Weakest Link.	J. Haigh.
STATISTICS IN MUSICOLOGY.	J. Beran.
GGE BIPLOT ANALYSIS: A GRAPHICAL TOOL FOR BREEDERS, GENETICISTS, AND AGRONOMISTS.	W. Yan and M.S. Kang.
KARL PEARSON: THE SCIENTIFIC LIFE IN A STATISTICAL AGE.	T.M. Porter.
SMALL WORLDS.	D.J. Watts.
DICING WITH DEATH.	S. Senn.
THE MATHEMATICAL CENTURY: THE 30 GREATEST PROBLEMS OF THE LAST 100 YEARS.	P. Odifreddi. With a Foreword by F. Dyson.
GAMMA EXPLORING EULER'S CONSTANT.	J. Havil. With a Foreword by F. Dyson.
PROBABILITY THEORY: THE LOGIC OF SCIENCE.	E.T. Jaynes. Edited by G.L. Bretthorst.
INTRODUCTION TO APPLIED STATISTICS: A MODELLING APPROACH. 2nd edition.	J.K. Lindsey.
ELEMENTS OF STOCHASTIC MODELLING.	K. Borovkov.
FUNDAMENTALS OF PROBABILITY AND STATISTICS FOR ENGINEERS.	T.T. Soong.
STATISTICS FOR EPIDEMIOLOGY.	N.P. Jewel.
MULTIPLE ANALYSES IN CLINICAL TRIALS. FUNDAMENTALS FOR INVESTIGATORS.	L.A. Moyé.
A FIRST COURSE IN STOCHASTIC MODELS.	H.C. Tijms.
ALL OF STATISTICS. A CONCISE COURSE IN STATISTICAL INFERENCE.	L. Wasserman.
MULTIVARIATE BAYESIAN STATISTICS: MODELS FOR SOURCE SEPARATION AND SIGNAL UNMIXING.	D.B. Rowe.
REGRESSION ANALYSIS. (A Constructive Critique).	R.A. Berk.
ASSOCIATION SCHEMES. DESIGNED EXPERIMENTS, ALGEBRA AND COMBINATORICS.	R.A. Bailey.
THE DESIGN AND ANALYSIS OF COMPUTER EXPERIMENTS.	T.J. Santner, B.J. Williams and W.I. Notz.
MODELLING THE INTERNET AND THE WEB: PROBABILISTIC METHODS AND ALGORITHMS.	P. Baldi, P. Frasconi and P. Smyth.
BAYESIAN ARTIFICIAL INTELLIGENCE.	K.B. Korb and A.E. Nicholson.
PROCRUSTES PROBLEMS.	J.C. Gower and G.B. Dijksterhuis.
DATA MINING: MULTIMEDIA, SOFT COMPUTING, AND BIOINFORMATICS.	S. Mitra and T. Acharya.
BAYESIAN REASONING IN DATA ANALYSIS: A CRITICAL INTRODUCTION.	G. D'Agostino.
STATISTICAL INFERENCE AND SIMULATION FOR SPATIAL POINT PROCESSES.	J. Møeller and R.P. Waagepetersen.
SPATIAL DATA ANALYSIS: THEORY AND PRACTICE.	R. Haining.
TEACHING STATISTICS USING BASEBALL.	J. Albert.
GRADE INFLATION – A CRISIS IN COLLEGE EDUCATION.	V.E. Johnson.
STATISTICAL SIZE DISTRIBUTIONS IN ECONOMICS AND ACTUARIAL SCIENCES.	C. Kleiber and S. Kotz.
EXTREME VALUES IN FINANCE, TELECOMMUNICATIONS, AND THE ENVIRONMENT.	B. Finkelstädt and H. Rootzén (Eds.).
FINANCIAL AND ACTUARIAL STATISTICS. AN INTRODUCTION.	D.S. Borowiak.
STATISTICAL ANALYSIS OF FINANCIAL DATA IN S-PLUS.	R. Carmona.
RESAMPLING METHODS FOR DEPENDENT DATA.	S.N. Lahiri.
SYSTEM RELIABILITY THEORY. MODELS, STATISTICAL METHODS AND APPLICATIONS, 2nd edition.	M. Rausand and A. Hoyland.
INTRODUCTION TO RARE EVENT SIMULATION.	J.A. Bucklew.
AUXILIARY SIGNAL DESIGN FOR FAILURE DETECTION.	S.L. Campbell and R. Nikoukhah.
INTRODUCTION TO STOCHASTIC SEARCH AND OPTIMIZATION.	J.C. Spall.
SENSITIVITY AND UNCERTAINTY ANALYSIS: THEORY, VOLUME I.	D.G. Cacuci.
PROBABILITY AND STATISTICS THE SCIENCE OF UNCERTAINTY.	M.I. Evans and J.S. Rosenthal.
APPLIED PROBABILITY.	K. Lange.
AN INTRODUCTION TO MULTIVARIATE STATISTICAL ANALYSIS, 3rd edition.	T.W. Anderson.
THE THIRD MAN OF THE DOUBLE HELIX.	M. Wilkins.
MEASURING INTELLIGENCE: FACTS AND FALLACIES.	D.J. Bartholomew.
STATISTIQUE. LA THÉORIE ET SES APPLICATIONS.	M. Lejeune.
COGWHEELS OF THE MIND. THE STORY OF VENN DIAGRAMS.	A.W.F. Edwards. Foreword by I. Stewart.
BEYOND REASON: 8 GREAT PROBLEMS THAT REVEAL THE LIMIT OF SCIENCE.	A.K. Dewdney.
A FIRST COURSE IN COMBINATORIAL OPTIMIZATION.	J. Lee.
STATISTICS AND THE EVALUATION OF EVIDENCE FOR FORENSIC SCIENTISTS, 2nd edition.	C.C. Aitken and F. Taroni.
ENVIRONMENTAL STATISTICS METHODS AND APPLICATIONS.	V. Barnett.
APPLIED BAYESIAN MODELING AND CAUSAL INFERENCE FROM INCOMPLETE-DATA PERSPECTIVES.	A. Gelman and X.-L. Meng (Eds.).
KENDALL'S ADVANCED THEORY OF STATISTICS, 2nd edition, Volume 2B: Bayesian Inference.	A. O'Hagan and J. Forster.
STATISTICAL MODELS.	A.C. Davison.
THE KERNAL METHOD OF TEST EQUATING.	A.A. von Davier, P.W. Holland and D.T. Thayer.
MULTIVARIATE t-DISTRIBUTIONS AND THEIR APPLICATIONS.	S. Kotz and S. Nadarajah.
AN INTRODUCTION TO MODERN NONPARAMETRIC STATISTICS.	J.J. Higgins.
NONPARAMETRIC AND SEMIPARAMETRIC MODELS.	W. Härdle, M. Müller, S. Sperlich and A. Werwatz.
INTRODUCTION TO REGRESSION ANALYSIS.	M.A. Golberg and H.A. Cho.
ANALYSIS OF VARIANCE FOR RANDOM MODELS: Volume I, Balanced Data: Theory, Methods, Applications and Data Analysis.	H. Sahai and M.M. Ojeda.
EXPLORING MULTIVARIATE DATA WITH THE FORWARD SEARCH.	A.C. Atkinson, M. Riani and A. Cerioli.
RANDOM GRAPHS FOR STATISTICAL PATTERN RECOGNITION.	D.J. Marchette.
AUTOMATIC NONUNIFORM RANDOM VARlATE GENERATION.	W. Hörmann, J. Leydold and G. Derflinger.
BIOSTATISTICS. A METHODOLOGY FOR THE HEALTH SCIENCES.	G. van Belle, L.D. Fisher, P.J. Heagerty and T. Lumley.
BAYESIAN APPROACHES TO CLINICAL TRIALS AND HEALTH-CARE EVALUATION.	D.J. Spiegelhalter, K.R. Abrams and J.P. Myles.
DISEASE MAPPING WITH WinBUGS AND MLwiN.	A.B. Lawson, W.J. Browne, and C.L. Vidal Rodeiro.
THE STATISTICAL EVALUATION OF MEDICAL TESTS FOR CLASSIFICATION AND PREDICTION.	M.S. Pepe.
STATISTICAL ESTIMATION OF EPIDEMIOLOGICAL RISK.	K.-J. Lui.
ANALYZING MULTIVARIATE DATA.	J. Lattin, J.D. Carroll and P.E. Green.
SAMPLE SURVEY THEORY. Some Pythagorean Perspectives.	P. Knottnerus.
INTRODUCTORY BIOSTATISTICS.	C.T. Le.
DIAGNOSTIC CHECKS IN TIME SERIES.	W.K. Li.
STATISTICS AND FINANCE, AN INTRODUCTION.	D. Ruppert.
AN INTRODUCTON TO FINANCIAL OPTION VALUATION.	D.J. Higham.
RISK AND FINANCIAL MANAGEMENT. MATHEMATICAL AND COMPUTATIONAL METHODS.	C.S. Tapiero.
FINANCIAL MODELLING WITH JUMP PROCESSES.	R. Cont and P. Tankov.
BIOSTATISTICAL GENETICS AND GENETIC EPIDEMIOLOGY.	R. Elston, J. Olson and L. Palmer (Eds.).
MEASUREMENT ERROR AND MISCLASSIFICATION IN STATISTICS AND EPIDEMIOLOGY.	P. Gustafson.
LOGIT MODELS FROM ECONOMICS AND OTHER FIELDS.	J.S. Cramer.
MEASURES OF INTEROBSERVER AGREEMENT.	M.M. Shoukri.
SURVIVAL ANALYSIS USING S.	M. Tableman and J.S. Kim. With a contribution from S. Portnoy.
RANDOMIZATION IN IN CLINICAL TRIALS: THEORY AND PRACTICE.	W.F. Rosenberger and J.M. Lachin.
MODELLING SURVIVAL DATA IN MEDICAL RESEARCH, 2nd edition.	D. Collett.
STATISTICAL METHODS FOR SURVIVAL DATA ANALYSIS, 3rd edition.	E.T. Lee and J.W. Wang.
SEMIPARAMETRIC REGRESSION.	D. Ruppert, M.P. Wand and R.J. Carroll.
NONPARAMETRIC STATISTICAL METHODS FOR COMPLETE AND CENSORED DATA.	M.M. Desu and D. Raghavarao.
NONPARAMETRIC GOODNESS-OF-FIT TESTING UNDER GAUSSIAN MODELS.	Y.I. Ingster and I.A. Suslina.
NONPARAMETRIC STATISTICAL INFERENCE, 4th edition, revised and expanded.	J.D. Gibbons and S. Chakraborti.
NUMERICAL METHODS FOR NONLINEAR ESTIMATING EQUATIONS.	C.G. Small and J. Wang.
BAYESIAN NONPARAMETRICS.	J.K. Ghosh and R.V. Ramamoorthi.
BAYESIAN DATA ANALYSIS, 2nd edition.	A. Gelman, J.B. Carlin, H.S. Stern and D.B. Rubin.

Contents:
1. Introduction
2. Simple descriptive techniques
3. Some time-series models
4. Fitting time-series models in the time domain
5. Forecasting
6. Stationary processes in the frequency domain
7. Spectral analysis
8. Bivariate processes
9. Linear systems
10. State-space models and the Kalman filter
11. Non-linear models
12. Multivariate time-series modelling
13. Some more advanced topics
14. Examples and practical advice

The author has succeeded in writing an accessible textbook which is wide-ranging, up-to-date and covering both theory and practice. Following his guideline that rigorous mathematics and practicality can go together, the author offers a wealth of applicable concepts and methods by which real life time series can be analyzed. The text offers a plethora of worked examples while the last section in each chapter contains exercises of different levels of difficulty. Its sixth (and final) edition (first edition 1975) should continue to hold the book's reputation on the market as one of the most accessible and popular textbooks on time series currently available.

Contents:
1. Introduction
2. Characteristics of time series
3. ARMA modelling and forecasting
4. Parametric nonlinear time series models
5. Nonparametric density estimation
6. Smoothing in time series
7. Spectral density estimation and its applications
8. Nonparametric models
9. Model validation
10. Nonlinear prediction

Readership: Academic (researchers and postgraduate students in statistics, economics, finance, business); industry (investment banking, insurance)

A couple of quotations from the Preface serve to convey the style and purpose of the book: "The aim of this book is to advocate those modern nonparametric techniques that have proven useful for analyzing real time-series data, and to provoke further research in both methodology and theory for nonparametric time-series analysis"; "We hope that this book will reflect the power of the integration of nonparametric and parametric approaches in analyzing time-series data."
The book is aimed at a broad readership, the prerequisites being just a grounding in probability (not measuretheory) and statistical methods. The more technical material (proofs of theorems, etc.) is generally relegated to "Complements" sections. Also, most chapters end with "Further Reading" or "Bibliographic Notes".
Chapter 1 gives some examples of time series: linear (white noise, AR, MA, etc.) and nonlinear (ARCH, threshold, nonparametric autoregressive). Chapter 2 covers stationarity, autocorrelation, spectral densities, the periodogram, longmemory processes and mixing conditions. Chapter 3 focuses on ARMA models (best linear prediction, maximum likelihood estimation, order determination, diagnostics, and linear forecasting). Chapter 4 covers threshold models, ARCH and GARCH, and bilinear models. Various aspects of kernel density estimation are discussed in Chapter 5, including windowing and whitening, bandwidth selection, boundaries, and asymptotics. In Chapter 6 smoothing is addressed, in both time and state domains, splines, and estimation of conditional densities. Spectral density estimation occupies Chapter 7 with material on tapering, kernel estimation and prewhitening, "automatic" methods, and tests for white noise. Chapter 8 addresses multivariate local polynomial regression, functionalcoefficient autoregressive models, adaptive versions, additive models, and conditional variance models. In Chapter 9 model validation is considered: generalized likelihood ratio tests, tests on spectral densities, autoregressive versus nonparametric models, and threshold versus varying-coefficient models. The last chapter, Chapter 10, covers nonlinear prediction, with material on characteristic features thereof, point and interval prediction, and predictive distributions.
This is a book that one can read as a beginner or as an expert. Although there are plenty of theorems, there are also plenty of numerical examples, with both real and simulated data, and lots of pictures and graphics (SPLUS-style). The topics are very fully explained and discussed, and there are many pointers to the literature for further study (with about six hundred references listed).

Contents:
PART A: Cases with Emphasis on Production Design
PART B: Cases with Emphasis on Development and Testing
PART C: Cases with Emphasis on Defect Prediction and Failure Analysis
PART D: Cases With Emphasis on Maintenance and Maintainability
PART E: Cases with Emphasis on Operations Optimization and Re-engineering
PART F: Cases with Emphasis on Product Warranty

Readership: Industry (manufacturing, maintenance, engineering); Academic (researchers and postgraduate students in Applied Statistics, Operational Research, Engineering)

This is an edited volume, with contributors from both academia and industry. There are applications from many branches of engineering giving a wide range of practical case studies. These are well set out, mostly beginning with a description of the context, the issues and objectives. This is followed by an outline of the mathematical and statistical methods to be used, after which the data are presented, discussed and analyzed, and conclusions drawn. A major aim, well served here, is to help bridge the gap between theory and practice.
The titles of the chapters are fairly self-explanatory. Each ends with a list of references and, unusually for an edited volume, exercises. Most chapters also present real sets of data. For these reasons, I would say that this book can make an extremely useful source for practitioners, students and teachers.

Contents:
1. Explorarory data mining and data cleaning: An overview
2. Exploratory data mining
3. Partitions and piecewise models
4. Data quality
5. Data quality: Techniques and algorithms

Readership: Data analysts who need to analyze large amounts of unfamiliar data, operations managers and students in undergraduate or graduate-level courses dealing with data analysis and data mining

The aim of the book is 'to develop a systematic process of data exploration and data quality management.' Up to about page 70, it is really an introduction to the sort of exploratory and descriptive statistical techniques used in data mining. After that, the practicalities of accessing and manipulating large sets of data are discussed, that is, from here on the differences between statistics and data mining become more apparent.
In Chapter 4, the book turns to a discussion of data quality. This has always been a key issue for data analysts; 'Garbage in, garbage out', as our computer scientist colleagues say. But with the growth in size of collections of data, the relevance of data-quality issues are becoming more apparent. The authors correctly point out that most data mining and data analysis books have assumed that the data have been cleaned prior to analysis, perhaps not surprisingly since the analytic stage is nearer to the excitement of the end result. Typically, rough and ready solutions have been adopted, such as dropping incomplete or distorted cases. Such a strategy in itself can lead to distortion, and it assumes that one knows that the data are distorted to start with.
There are some oddities in the book. For example, on p. 108, we find that 'Fisher (1966) is a good reference for the statistical design of experiments'. An important reference perhaps, but probably not one I would recommend to a newcomer to the area in the 21st century. And on p. 142, we find imputation defined as 'the process of guessing the values of missing data' .The running example is also rather disappointing, simply because it is artificial, albeit imaginative (an ecosystem containing Snarks, Gryphons and Unicorns). I am sorry, but I find that example queries such as (p. 75) 'For each year 1900 to 2002, report the average weight of Gryphons on the North face of Mt Everest,' just do not do it for me. I generally find that students are more convinced of the relevance of the tools I am describing when I use real examples. But maybe I am being too much of a purist.
The book does give many examples of how data can become corrupted, but, at the end, I wonder to what extent the problems one encounters in one's own studies are too specific to be helped by a general text such as this. This having been said, it is nice to see an attempt to inject some rigour and formality into this ubiquitous and very difficult problem.

Contents:
1. Introduction
2. Two simple data mining methods for variable assessment
3. Logistic regression: The workhorse of database response modeling
4. Ordinary regression: The workhorse of database profit modeling
5. CHAID for interpreting a logistic regression model
6. The importance of the regression coefficient
7. The predictive contribution coefficient: A measure of predictive importance
8. CHAID for specifying a model with interaction variables
9. Market segment classification modeling with logistic regression
10. CHAID as a method for filling in missing values
11. Identifying your best customers: Descriptive, predictive, and look-alike profiling
12. Assessment of database marketing models
13. Bootstrapping in database marketing: A new approach for validating models
14. Visualization of database models
15. Genetic modeling in database marketing: The GenIQ model
16. Finding the best variables for database marketing models
17. Interpretation of coefficient-free models

The back cover of this book says 'traditional statistical methods are limited in their ability to meet the modern challenge of mining large amounts of data' and then the contents lists a selection of predominantly traditional statistical methods. I wonder if we, the statisticians, are largely to blame for this misconception, by focusing, in our introductory courses, on detailed foundational aspects rather than on the more advanced tools which we teach in later courses.
This book is not aimed at statisticians. It includes very few mathematical formulae, which will make it attractive to its intended audience. On the other hand, I found that its simplifications grated considerably and could easily lead to misconceptions. For example, right on page 1 we read that 'mean profitability is not a valid summary measure if the individual profit values are not bell-shaped'. Apart from the puzzle of how individual values can be bell-shaped, the fact is that whether or not the mean is a valid summary measure depends on the question one is trying to answer.

Contents:
1. Introduction
PART I: Methodology
2. Organisation of the data
3. Exploratory data analysis
4. Computational data mining
5. Statistical data mining
6. Evaluation of data mining methods
PART II: Business Cases
7. Market basket analysis
8. Web clickstream analysis
9. Profiling web visitors
10. Customer relationship management
11. Credit scoring
12. Forecasting television audiences

Readership: Advanced undergraduate and postgraduate students of data mining, applied statistics, database management, computer science and economics

The subtitle shows that this book is restricted to commercial applications of data mining. Within those bounds, the book seeks to 'establish a bridge between data mining methods and applications'. It is thus intermediary between the (few) more rigorous methodological books on data mining and the (many) superficial books aimed at managers. The author claims that the first part 'gives a broad coverage of all methods currently used for data mining' – which, of course, is a challenge to reviewers to think of methods which are not covered. However, resisting such temptation, I will comment that the coverage is good, and certainly does include most of the main data mining tools. The first part of the book has chapters devoted to 'computational' and 'statistical' data mining. This distinction seems to me at best tenuous and at worst unsustainable in the modern world of data analysis. For example, the first of these chapters includes linear regression (surely a classic statistical tool if ever there was one), logistic regression (a type of glm, which is covered in the statistical chapter), and tree models (the seminal CART book was written by statisticians), while the second includes graphical models (an area developed in collaboration by computer scientists and statisticians). However, that is a cavil, and does not substantively detract from the value of the book. Its primary unique feature is the second part, which provides a set of case studies. l am sure these will be enlightening to anyone entering the area of data mining for the first time.
The book has a nice balance between theory and applications, and I will certainly recommend it to my students.

Contents:
1. The basics of credit risk management
2. Modeling correlated defaults
3. Asset value models
4. The credit risk model
5. Alternative risk measures and capital allocation
6. Term structure of default probability
7. Credit derivatives
8. Collateralized debt obligations

Readership: Risk managers and financial engineers looking for a quantitative approach to credit portfolio analysis. Students attendingpostgraduate courses in finance

This book is an introduction to quantitative credit risk. Its focus is primarily on statistical analysis of credit portfolios. After an introduction of the basic concepts, the most common credit portfolio models used in the industry (KMV, Credit Metrics, Credit Risk+) are described. The analysis is quantitative and is performed rigorously, without, however, loosing focus or clarity. Central is the description of the dependency structures between obligators.
In the second half of the book there is a brief overview of the most common credit derivatives. Pricing techniques are briefly considered to give just a flavour of the main issues. In the last chapter CDO and CLO structures are described. Even if only few characteristics of the product are considered, the chapter is interesting as it combines pricing with portfolio analysis.
This book can be useful particularly to credit risk managers, interested in quantitative analysis or implementation of credit portfolio models. I can therefore recommend it as a very nice and clear introduction to this topic.

Contents:
1. Introduction
2. Common thinking about risk and risk analysis
3. How to think about risk and risk analysis
4. How to assess uncertainties and specify probabilities
5. How to use risk analysis to support decision-making
6. Summary and conclusions

Readership: Anyone interested in the notion of risk, especially those with a less technical background in stochastics

Coming more from an engineering background, the author approaches risk and uncertainty from a predictive, Bayesian point of view. A risk analysis starts from prediction of observable quantities. Uncertainty with respect to the values of these quantities is expressed by means of probabilities. Models to be analyzed are deterministic functions of the observables. In the various chapters, this definition is worked out more precisely and illustrated on several examples. The approach is non-technical, at least for a probabilist or statistician. The latter may find the way in which someone outside the mainstream of stochastic-academia pedagogically handles the notion of risk.
I found the text refreshing, very well written, useful for its intended audience and in line with the way most statisticians would approach the problem.

Contents:
1. Introduction
2. Risk analysis methods
3. System definition and structure
4. Reliability assessment
5. Failure consequences and severity
6. Engineering economics and finance
7. Risk control methods
8. Data for risk studies

Readership: Statisticians, engineers, economists, finance people, reliability specialists

The book develops a philosophical foundation for the essentials of knowledge and ignorance. After offering the terminology and practice of risk management, it guides the reader to practical problems from engineering and economics. While the stochastic modelling component of risk analysis is only marginally treated, the book concentrates on algorithmic decision making and practical, conventional statistical procedures, common in reliability and financial engineering. A final chapter deals with information on data sources and failure.

Contents:
1. Foundations of financial risk management
2. Advanced analysis of financial risks
3. Insurance risks, foundations of actuarial analysis

Readership: Students interested in mathematical finance and insurance risk theory

An alternative (though somewhat long) title could have been: "An introduction to discrete time mathematical finance with an excursion to the continuous time world, a first encounter with the ruin model for insurance risk and examples of the interplay between insurance and finance." The first half of the book gives a readable introduction to what we now may call "classical discrete time mathematical finance". The standard results are complemented by more recent results on incompleteness and hedging in such markets. Several worked examples make the formulae come to life. The continuous time (Black-Scholes-Merton) model is obtained through a limit procedure and more formally through SDE theory. Chapter 3 introduces some basic insurance concepts, concentrating most on the famous Cramér-Lundberg type models where financial as well as insurance risk is present. Several examples, exercises, and C++ computer programs help the reader to digest the material introduced. The discussion of both insurance as well as finance models is a novelty. The way this combination is motivated may still look rather artificial from a practical point of view; it is however nice to see these processes treated (almost) at par in one text. This book may welI serve as a starting point for others to go the same road. It would have been nice if the publishers would have insisted on a final (extra) proof-reading, especially on the English.

Contents:
1. Weak convergence of stochastic processes
2. Weak convergence of financial markets
3. The basic models of approximation

The book recalls techniques and results of weak convergence of stochastic processes in mathematical finance and covers a wide range of applications. In the first chapter, results on weak convergence of stochastic processes are summarized; the second chapter deals with the question of how to apply these results, whereas in the third chapter techniques are given on how to construct discrete-time models which converge to continuous-time models. Most of the results presented are not new and are given without proof. For readers very well acquainted with the material, it may serve as a good reference book on the subject. A drawback is the sometimes uncommon numbering of sections, propositions, theorems etc. and the unsatisfactorily small subject index.

Contents:
1. Introduction and overview
2. Infinitely divisible distributions on the nonnegative integers
3. Infinitely divisible distributions on the nonnegative reals
4. Infinitely divisible distributions on the real line
5. Self-decomposability and stability
6. Infinite divisibility and mixtures
7. Infinite divisibility in stochastic processes

APPENDIX A: Prerequisites from Probability and Analysis
APPENDIX B: Selected Well-Known Distributions

A random variable is infinitely divisible if, for each positive integer n, it can be written (in distribution) as a sum of n independent random variables with the same distribution. This simply formulated concept has an enormous impact on the central limit theory for sums and on the theory of processes with stationary independent increments. But the concept also appears frequently in various more practical aspects of stochastic modelling. The two authors have been working on this book for many years and now they finally come up with a rather complete, very detailed and self-contained survey of the state of the art around the theory and applications of univariate infinite divisibility. All important properties are collected in the three separate chapters according to the support of the distribution (nonnegative integers, nonnegative reals, reals). In each of these chapters, the appropriate transform plays the key role (generating function, Laplace-Stieltjes transform, characteristic function). The next three chapters deal with special aspects of infinite divisibility: self-decomposibility and stability, mixing and its widespread appearance in various stochastic processes (queues, branching processes, renewal processes, shot noise processes, ...). An important plus for this book is the rich variety of examples throughout the whole text. The monograph will certainly become the standard reference work on this fundamental concept.

Contents:
1. Finite state space, a trial run
2. Moving to Euclidian space, the real thing
3. Itô's approach in the Euclidian setting
4. Further considerations
5. Itô's theory of stochastic integration
6. Applications of stochastic integration to Brownian motion
7. The Kunita-Watanabe extension
8. Stratonovich's theory

Itô processes, Itô's lemma and Itô SDEs appear not only in numerous textbooks on probability and stochastic processes, but also in many more books on applications to fields such as engineering, economics and biology. Few if any of these more applied texts care about the deeper "reasons" and "culture" of Itô's fundamental contributions. Itô introduced his by now famous theory of stochastic differential equations to explain the Kolmogorov-Feller theory of Markov processes (hence also the title of this book). Hence understanding the structure of Markov processes was Itô's primary goal. In the first part of the book, the author nicely explains the historical developments underlying Itô's work on Markov processes. The second part then concentrates on stochastic integration as developed by Itô. The two final chapters present the Kunita-Watanabe extension to semi-martingales and Stratonovich's extension of Itô's integration theory. I consider this book a must read for anyone interested in the mathematical culture underlying Kiyosi Itô's path-breaking contributions to stochastic calculus. The author also presents a mathematical bridge to current work in the field. As such, this text offers the ideal basis for a graduate course in the subject.

Contents:
1. What is probability?
2. Lotteries
3. Football Pools, Premium Bonds
4. One coin, many games
5. Dice
6. Games with few choices
7. Waiting, waiting, waiting
8. Let's play best of three
9. TV games
10. Casino games
11. Bookies, the Tote, spread betting
12. This sporting life
13. Lucky for some – miscellanea
14. Probability for lawyers

APPENDIX I: Counting
APPENDIX II: Probability
APPENDIX III: Averages and Variability
APPENDIX IV: Goodness-of-fit Tests
APPENDIX V: The Kelly Strategy

This is the second edition of the book that was originally published in 1999 [Short Book Reviews, Vol. 19, p. 4]. Similar to the first edition, the book is a gambling or games manual that uses probability arguments to obtain good strategies of play. Several chapters have been updated since the first edition, many remain essentially the same as before, and a new chapter has been added. As well, there are appendices that contain the mathematical detail necessary to the probability and statistical methods used within the text. The book remains a good read and is truly a modern-day Hoyle in the original sense of Hoyle's books. Most of those who have emulated Hoyle merely provide the rules of play. Hoyle's original Short Treatise on the Game of Whist from the eighteenth century contains, in addition to the rules, some strategies of play for that game based on the theory of probability. The current book is written in that same spirit. For teachers, the book contains a wealth of examples that could be used in the classroom to illustrate elementary probability theory. The only drawback to the new edition is that the new chapter, "Probability for lawyers", although interesting and informative, does not seem to fit into the same spirit of the first edition unless one considers taking a case before the bar as a kind of game to be played.

Contents:
1. Some mathematical foundations of music
2. Exploratory data mining in musical spaces
3. Global measures of structure and randomness
4. Time series analysis
5. Hierarchical methods
6. Markov chains and hidden Markov models
7. Circular statistics
8. Principal component analysis
9 .Discriminant analysis
10. Cluster analysis
11. Multidimensional scaling

Readership: Those with a serious interest in the quantitative analysis of music who could cope with the material of two university years of mathematics and statistics

This earnest book presents a survey of statistical methods that can be applied to music. It is complete: beginning with the definition of groups on page 7 and related theorems beginning on page 10. By page 26, medians are defined proceeding through to wavelets. Readers should not be deterred by the completeness. Much can be learned without spending time on the mathematics. The introduction of a large number of basic and advanced statistical methods is well illustrated with graphical displays based on musical scores and performances. The selection of the music and the methods is quite broad and interestingly eclectic. The author's evident enthusiasm for the subject matter makes the reading most pleasurable. This book will be useful in expanding the methodology applied in the field.

Contents:
PART I: Genotype-by-Environment Interaction and Stability Analysis
1. Genotype-by-environment interaction
2. StabiIity analyses in plant breeding and performance trials
PART ll: GGE Biplot and Multi-Environment Trial Data Analysis
3. Theory of biplot
4. Introduction to GGE biplot
5. Biplot analysis of multi-environment trial data
PART III: GGE Biplot Software and Applications to Other Types of Two-Way Data
6. GGE biplot software – The solution for GGE biplot analyses
7. Cultivar evaluation based on multiple traits
8. QTL identification using GGE biplot
9. Biplot analysis of diallel data
10. Biplot analysis of host genotype-by-pathogen strain interactions
11. Biplot analysis to detect synergism between genotypes of different species

Readership: Plant and animal breeders, geneticists, agronomists, and statisticians interested in applications in these fields

A GGE biplot is a graphical display of principal components associated with Genotype main effects (G) and Genotype-by-Environment interactions (GE). Its aim is to facilitate the comparison of different genetic strains in different environments. The book describes the GGE methodology, promotes the authors' software program that carries out such calculations, and provides an impressive collection of examples of data analysis in breeding experiments.
The methodology is statistically straightforward and can be used with any two-way data. The authors' contribution has been to make the basics of multivariate analysis more palatable to plant and animal breeders. The methodology seems to have been well received by these users. The writing style is very clear, but this reviewer found the authors' occasional self-congratulatory outbursts to be unbecoming in a textbook.
Potential buyers need to know that the book is literally a companion to the software marketed by the authors via their website. Though most statisticians could carry out such analyses using software they already have, non-statisticians will be at a loss unless they purchase the software as well as the book.
Statisticians will be interested to see how simple statistical ideas, such as main effects and interactions, can help illuminate the many trade-offs that are involved in breeding decisions.

Contents:
1. Introduction
2. Lehrjahre of a poetic wrangler
3. Apostle of renunciation
4. Pearson's progress
5. Cultural historian of a political age
6. Intellectual love and the women question
7. Ether squirts and the inaccessibility of nature
8. Scientific education and graphical statistics
9. The statistical reformation
10. Epilogue: Composing a life

Karl Pearson's name remains familiar as a statistical pioneer and in particular as the inventor of the chi-squared goodness-of-fit test. This biography recognizes that he came to see as his life's major task the propagation of universal quantification and of the importance of statistical thinking throughout science and beyond. Most of the book, however, is concerned not with statistical technicalities but with K.P.'s personality and his philosophical, political and ethical thought.
The book arises from the massive scholarly project of absorbing the very large amount of material about K.P., including his voluminous correspondence. Many fascinating details are covered. A lonely close to tragic figure emerges, an intolerant man of desperate seriousness, deeply involved in the main preoccupations of left-wing thinking of the late 19th and early 20th century and yet never part of organized activity such as the Fabian Society. He saw scientific research as an ethical activity and, for example, deplored what he saw as the professionalization of science, i.e. the notion that scientific research is a career rather than a calling, replacing dedication to the Church in earlier epochs. Some of his views, such as his socialism and his strong support for careers for women, may be seen now with sympathy. Others of his views, such as his attitude to war as part of a process of selection of the fittest, to Empire and to racial issues are now distasteful, although one must recognize that his views were not unusual at the time.
The book is, I believe, an important addition to the history not only of statistics, but of an exceptionally interesting period of intellectual life. The book covers so much that it is perhaps unfair to ask for more but I would like to have seen further discussion of his attitude to and the effect on him of the Great War. As even superficial perusal of Biometrika will show, the period 1900 to 1914 was for K.P. one of intense productive and extremely wide-ranging activity; although he wrote extensively from 1918 to his death little of it seems as impressive as the early work. Dr. Porter notes that K.P.'s early enthusiasm for German philosophy and culture had waned somewhat by 1914, particularly because of the dominance of Prussia; even so his commitment to German thought, as with many European intellectuals of the time, presumably remained considerable.
The final question that must be asked of any biography, especially one based solely on correspondence and indirect knowledge, is: was the subject really like this? Dr. Porter's care and thoroughness make a strong case that he has indeed described close to the real man!

Contents:
1. Kevin Bacon, the small world, and why it all matters
2. An overview of the small-world phenomenon
3. Big worlds and small worlds: Models of graphs
4. Explanations and ruminations
5. "It's a small world after all": Three real graphs
6. The speed of infectious disease in structured populations
7. Global computation in cellular automata
8. Cooperation in a small world: Games on graphs
9. Global synchrony in populations of coupled phase oscillators
10. Conclusions

This book (which appeared first in hardback form in 1999) is an excellent narrative describing work on network connectivity, which the author initiated with his Ph.D. supervisor around 1995. Their joint paper ("Collective dynamics of 'small-world' networks" Watts, D.J. and Strogatz, S.H. (1998) Nature 393, 440-442) has since stimulated an explosion of interest and hundreds of papers and articles. A good reason for the explosion is the tremendous range of networks to which the ideas can be applied: in communication, disease, transport, and so on.
The 'small world phenomenon' originated much earlier, of course ? empirically in work by the psychologist Stanley Milgram in the 1960's and more recently through the play 'Six Degrees of Separation' (1990) by John Guare. The popular/cult 'Kevin Bacon Game' involves connecting actors and actresses stepwise through other actors and actresses with whom they have appeared in movies.
The developing field involves the statistics of networks and, for example, their robustness, for example, How vulnerable would the WEB be to selective attack?
Finally, it is worth noting that these important and ongoing studies originated with curiosity-driven research. Watts's Ph.D. dissertation was about the synchronization of biological oscillators, in particular a population of chirping crickets.

Contents:
1. Circling the square
2. The diceman cometh
3. Trials of life
4. Of dice and men
5. Sex and the single patient
6. A hale view of pills
7. Time's tables
8. A dip in the pool
9. The things that bug us
10. The law is an ass
11. The empire of the sun

This book is designed to address the question: How do we translate information into knowledge? Statistics enables us to evaluate evidence, to design proper experiments and to formulate decisions effectively. These ideas are fundamental in all walks of life.
The author's interest is primarily in statistics as applied to medicine ? care, allocation of resources, risk, clinical trials, and so on. The aim is to entertain and the wealth of examples presented does indeed do this, while getting across sound methodology and a wide perspective of the historical development of statistical ideas.
The style of the book is discursive and it is very entertaining, although I suppose that is not for all tastes! In addition to being excellent for students and practitioners of the subject, it should be made essential reading for all those in public life who make critical decisions in the areas of medicine, politics, law and the media.

Introduction
1. The foundations
2. Pure mathematics
3. Applied mathematics
4. Mathematics and the computer
5. Open problems

This is a rather enjoyable account of the "peak episodes" in the mathematics of the last century. As a starting point, the author takes up the famous Hilbert problems presented by David Hilbert at the 1900 Paris Congress of Mathematicians. In his scenic tour of the mathematical century, he touches upon diverse aspects of mathematics, ranging from Wiles' proof of Fermat's last theorem, the classification of finite simple groups, the four colour theorem, knot theory, complexity theory and artificial intelligence. In the final chapter, he lists a few open problems, some that are now listed in the 2000 Clay Institute Millenium problems, the most notorious, being of course, the Riemann hypothesis. The book contains a delightful introduction by Freeman Dyson.

Introduction
1. The logarithmic cradle
2. The harmonic series
3. Sub-harmonic series
4. Zeta functions
5. Gamma's birthplace
6. The gamma function
7. Euler's wonderful identity
8. A promise fulfilled
9. What is gamma … exactly?
10. Gamma as a decimal
11. Gamma as a fraction
12. Where is gamma?
13. It's a harmonic world
14. It's a logarithmic world
15. Problems with primes
16. The Riemann initiative

which is numerically equal to 0.577215664901... We still do not know if this number is rational or irrational. Its importance emanates from the fact that it arises in the study of many functions, such as the Gamma function of Euler interpolating the factorials, and the zeta function of Riemann, to cite two cases. This book is a popular mathematics account of the history of this constant. Using ? as a central theme, the book discusses in a lively fashion, some of the major open problems of mathematics, such as the Riemann hypothesis.

Editor's Foreword
PART I: Principles and Elementary Applications
1. Plausible reasoning
2. The quantitative rules
3. Elementary sampling theory
4. Elementary hypothesis testing
5. Queer uses for probability theory
6. Elementary parameter estimation
7. The central Gaussian, or normal distribution
8. Sufficiency, ancillarity, and all that
9. Repetitive experiments: Probability and frequency
10. Physics of 'random experiments'
PART II: Advanced Applications
11. Discrete prior probabilities: The entropy principle
12. Ignorance priors and transformation groups
13. Decision theory: Historical background
14. Simple applications of decision theory
15. Paradoxes of probability theory
16. Orthodox methods: Historical background
17. Principles and pathology of orthodox statistics
18. The Ap distribution and rule of succession
19. Physical measurements
20. Model comparison
21. Outliers and robustness
22. Introduction to communication theory

APPENDIX A: Other Approaches to Probability Theory
APPENDIX B: Mathematical Formalities and Style
APPENDIX C: Convolutions and Cumulants

Readership: Scientists who have to make inference from incomplete information, graduates studying data analysis. No previous knowledge of probability or statistics is needed

The theme of this book is 'probability theory as extended logic', and it includes all Bayesian and all frequentist calculations as special cases of its rules, although, Jaynes argues, frequentist methods are 'quite inadequate for the current needs of science'. Jaynes was an outspoken proponent of Bayesian methods during the frequentist/Bayesian debates but here he bases his arguments not on philosophical or ideological arguments, but on 'facts of actual performance', and explains his own brand of the Bayesian school. This differs from, for example, that of de Finetti ('little more than a loose philosophical agreement remains; on many technical issues we disagree strongly with de Finetti. It appears to us that his way of treating infinite sets has opened up a Pandora's box of useless and unnecessary paradoxes', p. xxi).
E.T. Jaynes died in 1998, before this book was finished. Prior to his death, he had asked G. Larry Bretthorst to finish and publish it. Since there were missing chapters and missing parts of chapters, Bretthorst had to decide whether to create replacements or leave them missing. He chose the latter strategy, so that the material is pure Jaynes. As always with Jaynes's writing, it is lucid and stimulating.
This book will be illuminating for statisticians, philosophers, and everyone interested in inductive inference. I cannot resist suggesting that it should also be a must for fuzzy logicians and fuzzy set theorists. Despite its length, this book would amply repay careful study. It would be one of the ten I would choose to have with me were I to be shipwrecked on a desert island.

Contents:
1. Basic concepts
2. Categorical data
3. Inference
4. Probability distributions
5. Normal regression and ANOVA
6. Dependent responses
7. Where to now?

Readership: Students in medicine, biology, the social sciences, and applied statistics

This is a fairly traditional introductory applied statistics text, adequate for a 'two semester course of about 60 hours of lectures, plus practical tutorial work'. It requires the students to get their hands dirty, and seeks to motivate them about the aims and power of statistics. I agree with the author's opinion, stated in the preface to the first edition, that 'the approach presented here should be used as a first introduction to statistics even for the most mathematically sophisticated students. The feel for the aims of statistics should be clearly communicated before elaborate mathematical justifications are presented.'
The text has been partially reordered, with fundamental concepts being introduced in Chapter 1, with more explanation in places, and with the examples being separated from the text. Apart from in Chapter 4, little new material has been added.
I recommend it as a text for a basic course in applied statistics, either as a first statistics course to mathematics students, or as a more advanced course to students in other disciplines.

Contents:
1. Introduction
2. Basics of probability theory
3. Markov chains
4. Markov decision processes
5. The exponential distribution and Poisson process
6. Jump Markov processes
7. Elements of queueing theory
8. Elements of renewal theory
9. Elements of time series
10. Elements of simulation

Readership: Undergraduates with a reasonable knowledge of a first-year real analysis course

The decline in mathematical knowledge amongst students entering university (coupled with the very pervasive perception that mathematics is difficult and any subject involving it is best avoided) makes the teaching of rigorous courses in probability theory and stochastic processes, at least to classes of a size that university administrations regard as viable, virtually impossible. The author comes from the former Soviet Union where rigorous undergraduate courses in these subjects were standard fare. Since moving to Australia, he has encountered the realities that many of us are grappling with ? students with generally weak mathematical preparation. The text represents his response to this particular challenge and he has managed to strike a nice balance between the desire for a formal logical development of the subject and the need to maintain interest, to keep the material accessible and to include some straightforward examples and exercises (with solutions ? a must these days) that ensure that the less able student is not alienated. Some skill is needed to chart a course through a modern approach to stochastic modelling that avoids deeper mathematical techniques. There have been quite a few books at an elementary level that introduce these subjects whilst minimizing the need to prove results and this text too avoids proofs wherever possible. As acknowledged by the author, such an approach can have the drawback that a mathematically able student is not really stretched by the material. The author handles this by including plenty of links, at appropriate points throughout the text, to more sophisticated sources that do cover the material thoroughly as well as including sections in small font that provide details of interest to the more sophisticated reader but which are not essential for the general overview. There are also numerous exercises at the end of each chapter and the text is leavened with frequent footnotes containing relevant biographical material and interesting anecdotes. In a sense this book is a closer approximation to a rigorous approach than many introductory texts whilst simultaneously remaining accessible.

Contents:
1. Introduction
PART A: Probability and Random Variables
2. Basic probability concepts
3. Random variables and probability distributions
4. Expectations and moments
5. Functions of random variables
6. Some important discrete distributions
7. Some important continuous distributions
PART B: Statistical Inference, Parameter Estimation, and Model Verification
8. Observed data and graphical representation
9. Parameter estimation
10. Model verification
11. Linear models and linear regression

In the Preface the author says that 'As an introductory course, a sound and rigorous treatment of the basic principles is imperative. …' This gives the flavour of the book. Thus, the presentation verges towards the theoretical rather than the practical in many places, e.g. the Cramer-Rao Bound is given, together with proof, as is the central limit theorem, and the Neyman-Fisher Factorisation Theorem is stated.
Chapters 2 to 7 are on probability, and Chapter 8 contains a few pages on histograms. The statistical coverage, which is given in the last three chapters, comprises estimation (Chapter 8), goodness-of-fit (chi-square and Kolmogorov-Smirnov tests in Chapter 9), and linear regression (Chapter 10). Several statistical topics that would normally be thought appropriate for an applied course for engineers are omitted; for instance, examples of confidence intervals and parametric hypothesis tests (not just in connection with linear regression), binary and categorical data (logit regression and contingency tables), Anova, time series, etc.

Contents:
1. Introduction
2. Measures of disease occurrence
3. The role of probability in observational studies
4. Measures of disease-exposure association
5. Study designs
6. Assessing significance in a 2 x 2 tabIe
7. Estimation and inference for measures of association
8. Causal inference and extraneous factors: Confounding and interaction
9. Control of extraneous factors
10. Interaction
11. Exposures at several discrete levels
12. Regression models relating exposure to disease
13. Estimation of logistic regression model parameters
14. Confounding and interaction within logistic regression models
15. Goodness-of-fit tests for logistic regression models and model building
16. Matched studies
17. Alternatives and extensions to the logistic regression model
18. Epilogue: The examples

Readership: Public health researchers who have had a course in statistics that covers random variables, sampling, estimation and confidence intervals. Readers should be comfortable reading formulae.

Using examples, this experienced statistician identifies scientific issues and clearly links them to statis-tical approaches. Statistical theory and formality are grounded in manageable yet realistic examples. Coverage includes the basics and important topics such as measurement error and causal analysis. The book has excellent references, an informative index and glossary.

Contents:
0. Blossoms on a healthy plant
1. Fundamentals of a clinical trial design
2. Multiple analyses and the random experiment
3. The lure and complexity of multiple analyses
4. Multiple analyses and multiple endpoints
5. Introduction to multiple dependent analyses I
6. Multiple dependent analyses II
7. Introduction to composite endpoints
8. Multiple analyses and composite endpoints
9. Introduction to subgroup analyses
10. Subgroups II: Effect domination and controversy
11. Subgroups III: Confirmatory analyses
12. Multiple analyses and multiple treatment arms
13. Combining multiple analyses
14. Conclusions: The two-front war

APPENDIX A: Case Reports and Causality
APPENDIX B: Estimation in Random Research
APPENDIX C: Relevant Code of Federal Regulations
APPENDIX D: Sample Size Primer
APPENDIX E: Additional Dependent Hypothesis Testing Results

Readership: CIinical investigators at aIl levels and non-statisticians more generally. However, visiting the book is most definitely worthwhile for statisticians and epidemiologists

The book is entertaining and informative, sufficiently informal to recruit and retain the intended non-statistical readership, but sufficiently formal to detail methods. The author effectively sets up each issue with examples and conceptual discussion, grounding all in realistic examples. Material somewhat ancillary to multiple analyses (e.g., "Case reports and causality") broadens the book's base. But, the absence of even a mention of Bayesian approaches denies readers information on this effective approach to multiplicity and its burgeoning role in biomedical studies more generally.

Contents:
1. The Poisson process and related processes
2. Renewal-reward processes
3. Discrete-time Markov chains
4. Continuous-time Markov chains
5. Markov chains and queues
6. Discrete-time Markov decision processes
7. Semi-Markov decision processes
8. Advanced renewal theory
9. Algorithmic analysis of queueing models

APPENDIX A: Useful Tools in Applied Probability
APPENDIX B: Useful Probability Distributions
APPENDIX C: Generating Functions
APPENDIX D: The Discrete Fast Fourier Transform
APPENDIX E: Laplace Transform Theory
APPENDIX F: Numerical Laplace Inversion
APPENDIX G: The Root-finding Problem

This book is intended as "a first introduction to stochastic models for senior undergraduate students in computer science, engineering, statistics and operations research", and readers are assumed to be familiar with elementary probability theory. It would be suitable for mathematics or statistics undergraduates in the United Kingdom taking an introductory applied probability course in their second or third year, although there is far more material here than would be covered in a single course. The approach is generally clear and straightforward, with specific aspects of more complex models used to illustrate simpler processes. There are plenty of worked (OR-oriented) examples as well as a substantial set of exercises (without solutions) at the end of each chapter. Bibliographic notes and references are also given on a chapter by chapter basis. The author claims that applied probability teaching needs a fresh approach to take account of the computational developments of the last twenty years. Thus, the discussion of algorithms for the numerical solution of model properties is a useful, if generally low-key, feature of the book, and the Metropolis-Hastings algorithm and Gibbs sampler are included as applications of Markov chain theory. Nevertheless, the emphasis is on basic theory, with heuristic explanation and rigorous proofs or references as appropriate.
I do have some reservations though. For example, the concept of reversibility is not introduced until page 193 and is barely used, yet this important concept greatly simplifies the determination of equilibrium properties of stationary processes. More fundamentally, having often struggled to convey to students the conditional nature of the Markov property, I found the statement (p. 82) that "the future probabilistic behaviour of the process depends only on the present state of the process and is not influenced by its past history" to be particularly unhelpful.

Contents:
1. Probability
2. Random variables
3. Expectation
4. Inequalities
5. Convergence of random variables
6. Models, statistical inference and learning
7. Estimating the CDF and statistical functionals
8. The bootstrap
9. Parametric inference
10. Hypothesis testing and p-values
11. Bayesian inference
12. Statistical decision theory
13. Linear and logistic regression
14. Multivariate models
15. Inference about independence
16. Causal inference
17. Directed graphs and conditional acceptance
18. Undirected graphs
19. Loglinear models
20. Nonparametric curve estimation
21. Smoothing using orthogonal functions
22. Classification
23. Probability redux: Stochastic processes
24. Simulation methods

Readership: "...graduate or advanced undergraduates in computer science, mathematics, statistics, and related disciplines" (Preface)

This book was written specifically to give students a quick but sound understanding of modern statistics, and its coverage is very wide. With an average of eighteen pages per chapter (plus twenty pages of bibliography), not much depth is achieved anywhere. Chapter 13, for example, has three straight-line-fit examples, a page on least squares and maximum likelihood, a page on properties, a page on prediction, two pages on multiple regression, five pages on selection procedures, two pages of logistic regression, seven references, an appendix on Akaike, and eleven exercises. On to the next chapter! The book is extremely well done, but how many teachers will be brave enough to adopt it as a class text? Few I believe, but because the book is excellent, I shall be happy to be proven wrong.

Contents:
1. Introduction
PART I: Fundamentals
2. Statistical distributions
3. Introductory Bayesian statistics
4. Prior distributions
5. Hyperparameter assessment
6. Bayesian estimation methods
7. Regression
PART II: Models
8. Bayesian regression
9. Bayesian factor analysis
10. Bayesian source separation
11. Unobservable and observable source separation
12. FMRI case study
PART III: Generalizations
13. Delayed sources and dynamic coefficients
14. Correlated observation and source vectors
15. Conclusion

APPENDIX A: FMRI Activation Determination
APPENDIX B: FMRI Hyperparameter Assessment

Readership: Teachers of "Classical Multivariate Statistics" or of "Multivariate Bayesian Statistics"; readers interested in a Bayesian approach to the "Source Separation Problem"

The author's objective for this book is to develop a Bayesian approach to the "Source Separation Problem" presenting statistical background as necessary. The canonical example is that of several simultaneous speakers at a cocktail party whose conversations are recorded at a number of microphones placed about the room; the problem is to separate each spoken source from the mixed signals delivered by the microphones. An application of interest is that of functional magnetic resonance imaging whereby the image of a person's brain might be observed to change in response to different stimuli.
The first half of the book is intended to supply the reader with statistical background sufficient to be able to begin the problem of "Bayesian Source Separation" in Chapter 10. It is recommended in the preface that "Those with sufficient breath [sic] and depth in the fundamental material may skip directly to Chapter 8 ..." ; this is good advice for all readers. The "fundamentals" seem hastily put together, contain fundamental flaws (for example likelihood defined to be a joint density and so precluding, for example, its use on censored data), needless non sequiturs (for example in the derivation of a vague prior for a variance parameter, p. 42), and equivocal explanations (for example as in the presentation and development of conjugate priors via the official sounding "Conjugate procedure" which proceeds by "writing down the likelihood, interchanging the roles of the random variable and the parameter, and 'enriching' the distribution so that it does not depend on the data"; inconsistently applied on pages 44-45). Together these defects render the statistical fundamentals unnecessarily abstruse to a novice.
Though an index exists, it is nearly useless. Where one might expect to find "likelihood" or "prior distribution" or even "Bayes' rule", one is startled to find instead such single concept entries as "ah ha", "between", "explicit", "hill", "increment", "large", and "moving", to name but a few. Personal favourites are "shift" and "shifts" (both of which appear reassuringly on the same page, 258) and "marginal" which, ironically, appears on nearly one third of the pages of the book. The index seems designed by a technology reminiscent of the very problem the book is meant to address.
And this brings us to the second, more substantive, half of the book. Here the text is based on the author's research, much of which has previously appeared only in technical reports; essentially no other authors are referenced. The models are based on multivariate normal distributions with conjugate priors and the mathematics seems straightforward. Estimation is via Gibbs sampling and the iterated conditional modes algorithm. The FMRl case study shows the complexity of the problem and its solution via this heavily computational approach; the unrealistic nature of the simplifying assumptions employed suggests that there is room for further research.
The book is likely to be of some interest to those working on the source separation problem; its value to others would appear to be minimal. A more mature work, aggressively edited, might produce an admirably thin volume such as those which were once the hallmark of the books produced by Chapman and Hall.

Contents:
1. Prologue: Regression analysis as problematic
2. A grounded introduction to regression analysis
3. Simple linear regression
4. Statistical inference for simple linear regression
5. Causal inference for the simple linear model
6. The formalities of multiple regression
7. Using and interpreting multiple regression
8. Some popular extensions of multiple regression
9. Some regression diagnostics
10. Further extensions of regression analysis
11. What to do

Wherever regression analysis is applied and practical conclusions are drawn, there are worries about the value of the data collected and whether the conclusions are valid. Such worries tend to be much higher in social science areas than in experimental science areas. This book assesses and tackles these problems, fleshing out a brief, skeleton course in regression analysis, and concentrating wholly on social science issues. This specialized book is Volume 11 in the publisher's "Advanced Quantitative Techniques in the Social Sciences Series".

Contents:
1. Association schemes
2. The Bose-Mesner algebra
3. Combining association schemes
4. Incomplete-block designs
5. Partial balance
6. Families of partitions
7. Designs for structured sets
8. Groups
9. Posets
10. Subschemes, quotients, duals and products
11. Association schemes on the same set
12. Where next?
13. History and references

Readership: Researchers in experimental design, pure mathematicians interested in applications of association schemes

As expected, Rosemary Bailey has presented us with a very scholarly work. She successfully draws the two threads of experimental design and the pure mathematics of association schemes together in this extremely thorough exposition. The mathematical concepts are explained lucidly and all the reader's potential queries are anticipated. There are large numbers of helpful diagrams to illustrate abstract concepts, and numerous examples and exercises.
Much of the material covered is available elsewhere but the advantage of this book is that it is all collected together. The author has been able to demonstrate links between different topics and clarify notation and definitions in an area where many terms such as "balance" and "orthogonal" are somewhat overused.
This book shows how the power of abstract algebra can have a considerable impact on our understanding of designed experiments. The author has made an important contribution to bridging the gap between these two areas.

Contents:
1. Physical experiments and computer experiments
2. Preliminaries
3. Predicting output from computer experiments
4. Additional topics in prediction methodology
5. Space-filling designs for computer experiments
6. Some criterion-based experimental designs
7. Sensitivity analysis, validation, and other issues

APPENDIX A: List of Notation
APPENDIX B: Mathematical Facts
APPENDIX C: PerK: Parametric Empirical Kriging

The intention of the book is to provide statistical tools for design of computer experiments; the subject has been evolving since the early 1970s and is very much associated with books by T. Naylor and J.P.C. Kljenen (at that era, computer-time was limited and an extremely valued commodity). From the mathematical point of view, the problem can be seen as an approximation of (computationally) complex models by simpler stochastic models. To do this, the authors almost completely resorted to the techniques rooted in design of spatial experiments and spatial sampling blended with Bayesian ideas. Some readers may find that "computer experiments" is just a good excuse to write about the latter.

Contents:
1. Mathematical background
2. Basic WWW technologies
3. Web graphs
4. Text analysis
5. Link analysis
6. Advanced crawling techniques
7. Modeling and understanding human behavior on the web
8. Commerce on the web: Models and applications

APPENDIX A: Mathematical Complements
APPENDIX B: List of Main Symbols and Abbreviations

Readership: Students, postdoctoral fellows, faculty members and researchers from a variety of disciplines, including computer science, machine learning, engineering, statistics, economics, business and the social sciences

This is a fascinating book. It describes an area of research to which the ideas of uncertainty, probability and statistics are essential. They are needed both because the subject matter is noisy and because it is dynamic: even as I am writing this review, the web is growing and evolving. Furthermore, the web provides an excellent illustration of the phenomenon of emergent properties: by its very nature, involving large numbers of interacting systems, it possesses aggregate properties not possessed by any of its elements. Statistical methods, again, are the natural tool for modelling and understanding such properties.
From a teaching perspective, one could certainly use this book to motivate statistics teaching to engineers, computer scientists, and the like. The relevance of statistics is brought home very powerfully. From a research perspective, the book makes it very clear that this is an ideal research area for an innovative young statistician seeking a topic which is of great and dramatically growing importance.
I am sure the mistake on page 23 was put there deliberately, in order to induce a warm glow of smugness in the reader.
I highly recommend the book.

Contents:
1. Bayesian reasoning
2. Introducing Bayesian networks
3. Inference in Bayesian networks
4. Decision networks
5. Applications of Bayesian networks
6. Learning linear causal models
7. Learning probabilities
8. Learning discrete causal structure
9. Knowledge engineering with Bayesian networks
10. Evaluation
11. KEBN case studies

Readership: Advanced undergraduates in computer sciences with some background in
artificial intelligence

For the authors, understanding Bayesian artificial intelligence is the incorporation of Bayesian inferential methods in the development of a software architecture for artificial intelligence. The central notion is (KEBN) Knowledge Engineering Bayesian Networks. The elements of Bayesian network technology, automated causal discovery and learning probabilities from data are presented, aiming at a practical and accessible introduction to the main concepts in the technology, while paying attention to fundamental issues.
To cite the authors: "Expecting Bayesian principles to answer all questions about reasoning is expecting too much. Nevertheless, we shall show that Bayesian principles implemented in computer programs can deliver a great deal more than the nay-sayers have ever delivered".

Contents:
1. Introduction
2. Initial transformations
3. Two-set Procrustes problems - generalities
4. Orthogonal Procrustes problems
5. Projection Procrustes problems
6. Oblique Procrustes problems
7. Other two-sets Procrustes problems
8. Weighting, scaling and missing values
9. Generalised Procrustes problems
10. Analysis of variance framework
11. Incorporating information on variables
12. Accuracy and stability
13. Links with other methods
14. Some application areas, future, and conclusion

APPENDIX A: Configurations
APPENDIX B: Rotations and Reflections
APPENDIX C: Orthogonal Projections
APPENDIX D: Oblique Axes
APPENDIX E: A Minimisation Problem
APPENDIX F: Symmetric Matrix Products

Procrustes analysis concerns the (linear) transformation of two or more matrices so as to optimize some criterion of "closeness" to each other. It first arose in Psychometry for matching matrices of factor loadings, was adopted by statisticians for matching matrices of co-ordinates arising from ordination or scaling, and has found applications in many areas including sensory analysis and shape analysis. Actual methodology is dependent upon the type of transformation (orthogonal, oblique, projective), optimality criterion (least-squares, inner product, group average), numerical algorithm, weighting or scaling, and number of matrices used in the comparison, so a plethora of disparate techniques now exists in the literature.
The authors pick their way expertly through this minefield. They systematically detail the many methods, draw out their similarities and differences, elucidate the computational algorithms, show how other techniques link to Procrustes analysis, and even sketch out some new areas of potential study. The emphasis is predominantly on the underlying linear algebra and theory, but sufficient examples are provided to give a flavour of the applications. This is the first comprehensive text on the topic, and will undoubtedly prove invaluable to many researchers.

Contents:
1. Introduction to data mining
2. Soft computing
3. Multimedia data compression
4. String matching
5. Classification in data mining
6. Clustering in data mining
7. Association rules
8. Rule mining with soft computing
9. Multimedia data mining
10. Bioinformatics: An application

Readership: Professionals, researchers, and graduate students in data mining, machine learning, information retrieval and artificial intelligence

This book describes a collection of data-mining tools for handling the uncertain situations which real-Iife problems present. As such, one might expect it to be solidly based on the science of uncertainty, namely probability theory. In fact, however, the book describes 'soft computing', 'a consortium of methodologies [that provides] flexible information processing capability for handling real-Iife ambiguous situations,' and which includes fuzzy logic, neural networks, genetic algorithms, rough sets, and signal-processing tools such as wavelets. Personally, I find this a rather eclectic collection, and am not entirely convinced that it forms a 'paradigm' as the authors claim. In particular, I believe that most statisticians take the view that probability provides an effective approach to handling uncertainty of all kinds, and that fuzzy theory is unnecessary: the authors' statement that 'the probability of an event is related to the number of times it occurs (i.e., its frequency)', suggests they are taking a narrow frequentist view. Partly because of this, I think that statisticians will be uncomfortable with much of the book.
Despite the curious choice of tools grouped here, the descriptions will provide accessible introductions for people new to them, though I think they will appeal more to computer scientists than statisticians.

Contents:
PART 1: Critical Review and Outline of the Bayesian Alternative
1. Uncertainty in physics and the usual methods of handling it
2. A probabilistic theory of measurement uncertainty
PART 2: A Bayesian Primer
3. Subjective probability and Bayes' theorem
4. Probability distributions (a concise reminder)
5. Bayesian inference of continuous quantities
6. Gaussian likelihood
7. Counting experiments
8. Bypassing Bayes' theorem for routine applications
9. Bayesian unfolding
PART 3: Further Comments, Examples and Applications
10. Miscellanea on general issues in probability and inference
11. Combination of experimental results: A closer look
12. Assymmetric uncertainties and nonlinear propagation
13. What priors for frontier physics?
PART 4: Conclusion
14. Conclusions and bibliography

This is an interesting book to read. It contains a number of non-standard thoughts, short philological essays and a good collection of carefully selected citations. Without any doubt probabilistic modeling (and consequently the Bayesian approach) is one of the simplest and straight-forward ways to quantify uncertainties, but not the only one (fuzzy set theory and interval mathematics are most popular examples). At least a short paragraph with comparative analysis would help a reader to better understand the place of the Bayesian ideas and the corresponding probability models in describing experiments in physics or any other science. At some places the presentation is less mathematically rigorous, but on the whole it is an exciting text.

Contents:
1. Examples of spatial point patterns
2. Introduction to point processes
3. Poisson point processes
4. Summary statistics
5. Cox processes
6. Markov point processes
7. Metropolis-Hastings algorithms
8. Simulation-based inference
9. Inference for Markov point processes
10. Inference for Cox processes
11. Birth-death processes and perfect simulation

APPENDIX A: History, Bibliography and Software
APPENDIX B: Measure Theoretical Details
APPENDIX C: Moment Measures and Palm Distributions
APPENDIX D: Perfect Simulation of SNCPs
APPENDIX E: Simulation of Gaussian Fields
APPENDIX F: Nearest-neighbour Markov Point Processes
APPENDIX G: Results for Spatial Birth-death Processes

Readership: Statisticians and applied probabilists, from postgraduate level upwards

Spatial point process models have applications in a wide variety of areas, and their formal theory is well developed. Inference for such models is, however, notoriously difficult except in a few special cases. With the advent of fast, cheap computers, simulation-based inference for spatial point processes has become feasible. This book provides an excellent and up-to-date review of developments in this area. It covers most, if not all, of the major classes of models, and discusses methods for their approximate and exact simulation. An introduction to the general framework of simulation-based inference is given, before discussing its application in the specific context of spatial point processes. Several examples are given to illustrate the techniques. The style is predominantly mathematical throughout, although some of the more abstract measure theoretic details are relegated to the extensive appendices.

Introduction
PART A: The Context for Spatial Analysis
1. Spatial data analysis: Scientific and policy context
2. The nature of spatial data
PART B: Spatial Data: Obtaining Data and Quality Issues
3. Obtaining spatial data through sampling
4. Data quality: Implications for spatial data analysis
PART C: The Exploratory Analysis of Spatial Data
5. Exploratory spatial data analysis: Conceptual models
6. Exploratory spatial data analysis: Visualization methods
7. Exploratory spatial data analysis: Numerical methods
PART D: Hypothesis Testing and Spatial Autocorrelation
8. Hypothesis testing in the presence of spatial dependence
PART E: Modelling SpatiaI Data
9. Models for the statistical analysis of spatial data
10. Statistical modelling of spatial variation: Descriptive modelling
11. Statistical modelling of spatial variation: Explanatory modelling

APPENDIX I: Software
APPENDIX II: Cambridgeshire Lung Cancer Data
APPENDIX III: Sheffield Burglary Data
APPENDIX IV: Children Excluded from School: Sheffield

Readership: Academic (undergraduate and MSc students in geography, social sciences, economics, environmental science, biology, humanities)

In the preliminary paragraphs, the author describes courses he has given on the subject and collaborations he has been involved with in applying the methodology. Also mentioned, is his earlier book, Spatial Data Analysis in the Social And Environmental Sciences, but it is not clear how much overlap there is with this one.
The coverage here is pretty extensive and the presentation is mainly in words with just a few symbols and equations. Therefore, readers with modest mathematical technique will be well able to cope. Different areas of application are described in some detail with many references. (In fact, the list at the end comprises thirty pages of references.) There is also much discussion of issues around the application of the statistical methods, for example scientific and policy implications, data sampling and quality.
After reading this book, the conscientious student should become very knowledgeable about the area without becoming bogged down in technicalities. This seems very appropriate for the intended readership.

Contents:
1. An introduction to baseball statistics
2. Exploring a single batch of baseball data
3. Comparing batches and standardization
4. Relationships between measurement variables
5. Introduction to probability using tabletop games
6. Probability distributions and baseball
7. Introduction to statistical inference
8. Topics in statistical inference
9. Modelling baseball using a Markov chain

Readership: Students of statistics who need this familiar and appealing motivation on which to base their studies. Sports enthusiasts who are keen to learn more about baseball performance and strategy by the use of statistics

This book is designed to use the vehicle of a game familiar to all those students in its main markets (USA/Canada, Japan, Puerto Rico, Cuba, ...) to motivate the main ideas of data handling and statistical analysis. Baseball generates data on every aspect of the game and much more of it than any other game with any claim to be mainstream – certainly much more even than cricket! That baseball has retained essentially the same main features since its inception, means that most of the categories of data are quite comparable over a period stretching back to the nineteenth century.
The author was co-author with J. Bennett of an earlier book [Curve Ball, 2001, Springer-Verlag (Copernicus)], which celebrated the uses of statistics in baseball. In this new book, the aim is to formalize statistics teaching/learning using baseball as a vehicle. In an American culture, this is very likely to be successful, but probably not elsewhere. After all, those of us in Europe who watch games beamed live from the US Major Leagues, necessarily do so at hours of the day which are (rightly) regarded as extremely unsocial! Hence even the rules, let alone the finer points, are not common currency. The general idea, of course, might well be adaptable in some respects to other games.

Contents:
1. Introduction
2. The DUET experiment
3. Grades and student evaluations of teaching
4. DUET analysis of grades and SETs
5. Validity of SETs
6. Grades and student course selection
7. Grading equity
8. Conclusions

Readership: University administrators, politicians, educationists, statisticians involved in the design of assessment schemes for teaching and of robust grading procedures

This book examines the phenomenon of grade inflation and its impact on post-secondary education particularly in the USA. This is a very thorough analysis of the ways in which variability in grading practice – across institutions, faculties/departments and individual courses – have consequences for the popularity of student choice of options, in student evaluations of teaching (SETs) and in a general degradation of academic standards.
The author was centrally involved with a student evaluation exercise carried out at Duke University in 1998/1999 [DUET-Duke Undergraduates Evaluate Teaching]. The idea was to find the effects that grades (anticipated and received) had on student course selection decisions and on their evaluations of courses and lecturers.
The detailed story of DUET and the discussions of issues of completion rates and non-response, causal effects of grading, standardization procedures and grade adjustment, are essential reading – particularly for all those who either impose, or have imposed upon them, evaluation schemes which are claimed to give objective measures of these things. The discussions are relevant to all post-secondary education systems.
Of course, the regarding of 'inflated' grades as true measures of attainment levels, often reflects a self-interest of students, academic faculty, departments, institutions and politicians, so that this practice has no coordinated natural opposition – at least in the short term. However, the effects are quite pernicious and all grading/eval-uations should really be carried out through 'properly designed experiments'. No doubt this dialogue should and will continue!

Contents:
1. Introduction
2. General principles
3. Pareto distributions
4. Lognormal distributions
5. Gamma-type size distributions
6. Beta-type size distributions
7. Miscellaneous size distributions

APPENDIX A: Biographies
APPENDIX B: Data on Size Distributions
APPENDIX C: Size Distributions

Readership: Economists, economic statisticians, econometricians, actuarial statisticians, statisticians interested in continuous distribution theory and applications

This book is predominantly concerned with distributions of income and wealth, but also considers distributions of size of firms, and of actuarial losses. The authors' objective was "marshaling and knitting together the immense body of information scattered in diverse sources in at least eight languages". The distributions of interest are all univariate and continuous (multivariate extensions are very briefly covered in one or two cases). Each chapter begins with a brief informal comment on its contents. The general style and layout of chapters is reminiscent of Johnson, Kotz and Balakrishnan's Continuous Univariate Distributions, Volumes 1 and 2, 2nd edition (Wiley, 1994, 1995) [Short Book Reviews, Vol. 15, p. 25, p. 43]; the mathematical level is similar but there is rather more descriptive material and slightly less mathematical detail. The authors have tried to avoid unnecessary overlap with the two previous volumes and to place the material in its economic and social contexts. A convenient feature is the appendix of short bibliographies of some of the leading contributers to the theory and development of these distributions (Pareto, Benini, Lorenz, Gini, Amoroso, D'Addario, Gibrat, Champernowne). By such means, the authors have gone a long way towards achieving their ideal that "a useful book on this subject matter should be interesting".

Contents:
1. Statistics of extremes, with applications in environment, insurance, and finance
2. The use and misuse of extreme value models in practice
3. Risk management and extreme value theory
4. Extremes in economics and the economics of extremes
5. Modeling dependence and tails of financial time series
6. Modeling data networks
7. Multivariate extremes

Readership: Statisticians, actuaries, scientists with an interest in extreme values

This is a collection of papers presented at the seminar on extreme-value theory held in Gothenburg in December 2001 but more instructional and cohesive than a typical collection of papers. The book includes excellent discussions of the applications of extreme-value theory to risk management, insurance, rainfall and data networks, as well as overview of univariate and multivariate extreme value theory. Arguably extremal events in the environment and the economy constitute a more visible threat than in the past and so the careful statistical study of such events is increasingly important. This book provides an excellent source both for the basic successes and the limitations of the current extreme-value methodology.

Contents:
1. Statistical concepts
2. Financial computational models
3. Deterministic status models
4. Future lifetime random variable
5. Future lifetime models and tables
6. Stochastic status models
7. Scenario and simulation testing
8. Further statistical considerations

The simple models used in finance and actuarial problems are well known. The book combines these with the statistical background for beginning students. The book will thus be a suitable introductory text more for beginning students and financial investigators, less for more advanced students in statistics or actuarial sciences.
Deterministic as well as stochastic models are dealt with together. Mathematical derivations are kept on a simple level. Unfortunately, the processor used for mathematical formulae is not of the usual standard in books in statistics or mathematics. The book contains also some applied problems or exercises without solutions, for self-study.

Contents:
1. Univariate exploratory data analysis
2. Multivariate data exploration
3. Parametric regression
4. Local and nonparametric regression
5. Time series models: AR, MA, ARMA, and all that
6. Multivariate time series, linear systems and Kalman filtering
7. Nonlinear time series: Models and simulation

Readership: Researchers, practitioners and students having at least two years of undergraduate probability and statistics with an interest in financial engineering

This is a timely book on modern data analysis with a difference: the examples and applications are predominantly taken from Financial Engineeering. The book avoids the complexities of continuous time Itô processes by treating only discrete time models, but covers many of the topics in a modern course in data analysis or computational statistics. Part I deals with exploratory data analysis, including nonparametric density estimation, quantiles, heavy-tailed distributions and copulas. Part II deals with regression, parametric and non-parametric including smoothing splines, kernel regression, and nonparametric option pricing. Part III includes time series analysis, linear and non-linear systems and the Kalman filter. The book also includes an elementary tutorial on the use of S-Plus as well as code so that the reader can replicate most of the analyses and graphs. This book will help fill a statistical gap in the otherwise heavily theoretical literature in mathematical finance.

Contents:
1. Scope of resampling methods for dependent data
2. Bootstrap methods
3. Properties of block bootstrap methods for the sample mean
4. Extensions and examples
5. Comparison of block bootstrap methods
6. Second-order properties
7. Empirical choice of the block size
8. Model-based bootstrap
9. Frequency domain bootstrap
10. Long range dependence
11. Bootstrapping heavy-tailed data and extremes
12. Resampling methods for spatiaI data
13. Appendices A, B

I found this a useful book that organizes many scattered results in a reasonably concise form. The author states that this book has two main audiences, so the first five chapters are a pedantic introduction aimed at graduate students and the last seven a research monograph aimed at researchers in statistics and econometrics. In my opinion, the first part might have heen slightly more pedantic and the latter more critical. The emphasis is on theoretical results and, for a book on a computationally-intensive process, has surprisingly few figures. Most graduate students would want to read the first eight chapters of Davison and Hinkley (1999) before starting this book.
Time series methods are largely restricted to ARMA procedures and raw (untapered) periodograms. In most of the data, I see, for example Thomson (2001), both of these are too badly biased to be bothered computing, so bootstrapping them would be rather pointless. The only mention of tapers seems to be on page 228, "This is a special case of Theorem I of Dahlhaus and Janas (1996) when no data-tapers are used.", with no mention that the use of tapers is critical, Brillinger (1981). There is no mention of multi-taper estimates nor that they jackknife easily, Thomson and Chave (1991).
In summary, I learned quite a bit from reading this book and consider it a good reference book for the mathematically inclined.

References
Brillinger, D.R. (1981). The key role of tapering in spectrum estimation. IEEE Trans. on Acoustics, Speech, and Signal Processing, 29, 1075-1076.
Dahlhaus, R. and Janas, D. (1996). A frequency domain bootstrap for ratio statistics in time series analysis. Annals of Statistics, 24, 1934-1963.
Davison, A.C. and Hinkley, D.V. (1999). Bootstrap Methods and their Application. Cambridge University Press. [Short Book Reviews, Vol. 18, p. 25]
Thomson, D.J. (2001). Multitaper analysis of nonstationary and nonlinear time series data. In W. Fitzgerald, R. Smith, A. Walden, and P. Young, editors, Nonlinear and Nonstationary Signal Processing, pages 317-394. Cambridge University Press.
Thomson, D.J. and Chave, A.D. (1991). Jackknifed error estimates for spectra, coherences, and transfer functions. In S. Haykin, editor, Advances in Spectrum Analysis and Array Processing, volume 1, chapter 2, pages 58-113. Prentice-Hall.

Contents:
1. Introduction
2. Failure models
3. Qualitative system analysis
4. Systems of independent components
5. Component importance
6. Dependent failures
7. Counting processes
8. Marker processes
9. Reliability of maintained systems
10. Reliability of safety systems
11. Life data analysis
12. Accelerated life testing
13. Bayesian reliability analysis
14. Reliability data sources

This book is a revised and expanded follow-up to the first edition [Short Book Reviews, Vol. 15, p. 44]. It provides a comprehensive treatment of system reliability theory. As was the first edition, it is especially good in its discussion of both qualitative and quantitative aspects of system reliability. The coverage of models and discussion of practical matters is excellent. Two new chapters on the reliability of maintained and safety systems are very welcome. The treatment of reliability data analysis is quite cursory, but several other books deal in depth with this topic. The present volume remains one of the best treatments of system reliability and will be a valuable resource for anyone working or teaching in this area.

Contents:
1. Random number generation
2. Stochastic models
3. Large deviation theory
4. Importance sampling
5. The large deviation theory of importance sampling estimators
6. Variance rate theory of conditional importance sampling estimators
7. The large deviations of bias point selection
8. Chernoff's bound and asymptotic expansions
9. Gaussian systems
10. Universal simulation estimators
11. Rare event simulation for level crossing and queueing models
12. Blind simulation
13. The (Over-Under) biasing problem in importance sampling
14. Tools and techniques for importance sampling

APPENDIX A: Convex Functions and Analysis
APPENDIX B: A Covering Lemma
APPENDIX C: Pseudo-Random Generation Programs

Readership: Systems engineers, graduate students in operations research or communications engineering, applied probabilists

This book deals with a special topic in simulation, the estimation of small probabilities via importance sampling using large deviations theory to design and/or analyze the change of measure. The aim has been to give an introduction with the large deviations theory kept at a fairly elementary level (Cramér's theorem in multidimensions and the Gärtner-Ellis theorem) as well as examples at a simple level (often just large deviations in the law of large numbers). The book may, therefore, serve as a bridge to the quite technical research literature and a help to get started with more challenging examples. One should be aware that little or no discussion is given of the limitations of the large deviations approach or of recent important developments in rare event simulation such as multilevel splitting (RESTART), the adaptive cross-entropy method and heavy tails.

Contents:
1. Introduction
2. Failure detection
3. MuItimodel formulation
4. Direct optimization formulations
5. Remaining problems and extensions
6. Scilab programs

Readership: Engineers and applied mathematicians having an interest in fault detection

Failure detection refers to the problem of determining if a system's behaviour has departed from its nominal or usual behaviour. There exists a considerable literature dealing with this topic which ranges from statistical based methods (for example likelihood ratio tests) to deterministic methods based on bounded errors. This book adopts a deterministic approach using set-valued estimation methods. The prime focus is on the problem of designing the inputs to the system so as to guarantee separation of the responses of the nominal system and faulted system. Some of the methods can be given statistical interpretations with fixed confidence levels but these connections are only briefly alluded to in the book.

Contents:
1. Stochastic search and optimization: Motivation and supporting results
2. Direct methods for stochastic search
3. Recursive estimation for linear models
4. Stochastic approximation for nonlinear root-finding
5. Stochastic gradient form of stochastic approximation
6. Stochastic approximation and the finite-difference method
7. Simultaneous perturbation stochastic approximation
8. Annealing type algorithms
9. Evolutionary Computation I: Genetic algorithms
10. Evolutionary Computation ll: General methods and theory
11. Reinforcement learning via temporal differences
12. Statistical methods for optimization in discrete problems
13. Model selection and statistical information
14. Simulation-based Optimization I: Regeneration, common random numbers, and selection methods
15. Simulation-based Optimization ll: Stochastic gradient and sample path methods
16. Markov chain Monte Carlo
17. Optimal design for experimental inputs

APPENDIX A: Selected Methods from Multivariate Analysis
APPENDIX B: Some Basic Tests in Statistics
APPENDIX C: Probability Theory and Convergence
APPENDIX D: Random Number Generation
APPENDIX E: Markov Processes

Readership: Advanced level undergraduates, graduate students and researchers in applied mathematics, operations research, mathematical economics and engineering

Stochastic search and optimization concerns algorithms for decision making in industry, academia and government, in circumstances when the measurements available to the algorithms are corrupted by noise and/or randomness is artificially inserted into the algorithms for better performance. The book, which is largely self-contained, provides easy access to a very broad, but related, collection of topics, from basic stochastic approximation algorithms for root finding and optimization to powerful Markov Chain Monte Carlo (MCMC) methods for computing statistical estimates of expectations, where conventional, analytical or multi-dimensional numerical integration techniques fail. Along the way, material is included on simulated annealing, evolutionary computation, re-enforcement learning and experiment design. The care taken by the author to motivate the analysis and provide perspective, and also the extensive literature review, are distinguishing features of this book and ones that will help greatly to extend its readership.

Introduction
1. Concepts of matrix and operator theory for sensitivity and uncertainty analysis
2. Concepts of probability theory for sensitivity and uncertainty analysis
3. Measurement errors and uncertainties: Basic concepts
4. Local sensitivity and uncertainty analysis of linear systems
5. Local sensitivity and uncertainty analysis of nonlinear systems
6. Global optimization and sensitivity analysis

Readership: Theoretical researchers in mathematics, computer science, control theory, electrical engineering, nuclear engineering and other areas of application; and advanced, mathematically oriented students in these fields

This is a systematic mathematical treatment of the theoretical underpinnings of sensitivity and uncertainty analysis. The author focuses on the Forward Sensitivity Analysis Procedure (FSAP), the Adjoint Sensitivity Analysis Procedure (ASAP), and the use of deterministic sensitivities in uncertainty analysis. The first three chapters are concise reviews of matrix and operator theory, probability theory and concepts of measurement error; this is an enjoyable review for those who already know these topics, but not intended for those learning the concepts for the first time. The remaining chapters discuss the FSAP and ASAP approaches for local analysis of linear and nonlinear systems respectively, and global analysis of complex systems, all with illustrative examples.
Considerable mathematical sophistication is expected of the reader. The author's prose is cogent and precise and is pleasant to read. Yet this is a rigorous theoretical book and is not aimed at practitioners despite the illustrative examples. Rather, this is a core addition to the reference literature for researchers in the field.

Contents:
1. Probability models
2. Random variables and distributions
3. Expectation
4. Sampling distribution and limits
5. Statistical inference
6. Likelihood inference
7. Bayesian inference
8. Optimal inferences
9. Model checking
10. Relationships among variables
11. Advanced topic – stochastic processes

Readership: Students who have studied one year of calculus at university level and seeking an introduction to probability and statistics

This book is an introductory text on probability and statistics that is heavily biased towards the mathematical concepts involved. Most of the mathematical details are included, but it is expected that students have a mastery of calculus before attempting to follow this text. The book can be used without an appropriate statistical package. However, for the maximum benefit to be obtained from this text, access to a computer package would be advantageous. All the computations in the text were carried out using Minitab.
The authors have structured the exercises to enable users to select the exercises which are applicable to them. The structures are called exercises, problems, challenges, computer exercises and computer problems. Exercises are suitable for all students and are there to give practice in applying the concepts. Problems require a more in-depth approach by the students where greater understanding of the concepts is needed. Challenges are onIy for students who have no difficuIty with exercises and problems. Computer exercises and problems are for students to do using appropriate statistical packages. Minitab is the package used and recommended by the authors. Although the authors have included a wealth of questions, there are no solutions provided.
This is a book to recommend to students who wish to acquire a sound mathematical foundation in probability and statistics.

Contents:
1. Basic notions of probability theory
2. Calculation of expectations
3. Convexity, optimization and inequalities
4. Combinatorics
5. Combinatorial optimization
6. Poisson processes
7. Discrete-time Markov chains
8. Continuous-time Markov chains
9. Branching processes
10. Martingales
11. Diffusion processes
12. Poisson approximation
13. Number theory

Readership: Probabilists, graduate students, scientists interested in probabilistic thinking

The author tries to offer to the scientific community at large an introduction to some of the most important aspects of applied probability. From the table of contents, it is clear that the author has chosen a very personal approach by including chapters on combinatorial optimization, Poisson approximation and number theory. On the other hand, this choice illustrates the beauty, utility and relevance of probabilistic thinking in a variety of scientific areas. In particular, pretty applications in computer science and genetics strenghten the overall message of this book, namely to give applied probability the attention it deserves.

Contents:
1. Introduction
2. The multivariate normal distribution
3. Estimation of the mean vector and covariance matrix
4. The distributions and uses of sample correlation coefficients
5. The generalized T2-statistic
6. Classification of observations
7. The distribution of the sample covariance matrix and the sample generalized variance
8. Testing the general linear hypothesis; multivariate analysis of variance
9. Testing independence of sets of variables
10. Testing hypotheses of equality of covariance matrices and equality of mean vectors and covariance matrices
11. Principal components
12. Canonical correlations and canonical variables
13. The distributions of charasteristic roots and vectors
14. Factor analysis
15. Patterns of dependence; Graphical models

The first edition of this text appeared in 1958, the second in 1984 [Short Book Reviews, Vol. 5, p. 3], and now comes the third after another nineteen years. It contains a new chapter on graphical models, new sections on elliptically contoured distributions at the end of most chapters, and some extra material on reduced rank regression. The additions on elliptically contoured distributions extend the normal-based theory on which the book is centred to non-normal situations, and the graphical models are certainly welcome. However, the dramatic advancements in computing technology that have changed the face of multivariate analysis over the past twenty years are barely mentioned. Consequently, chapters such as those dealing with classification and principal components now present a very limited picture, while aspects such as computationally-intensive inference are not touched on at all. Nevertheless, within the confines of parametric multivariate theory this book remains an authoritative work that can still be highly recommended.

Contents:
1.Distant Shores
2.Finding my feet
3.In a world at war
4.Randall's circus
5.Crystal genes
6.Go back to your microscopes
7.How does DNA keep its secrets?
8.The double helix
9.Living with the double helix
10.A broader view

Readership: Anyone interested in the origins of molecular biology and the modern bioinformatic, genomic, and protemic revolution, and in the history of a controversial episode in science

This is the autobiography of Maurice Wilkins, who shared the 1962 Nobel Prize in physiology or medicine with Francis Crick and James Watson for the discovery of the structure of DNA.
Much has been written about the role of Rosalind Franklin in the discovery of DNA. Wilkins and she did not always get on very well, and there has been controversy about the way her contribution was treated. Wilkins writes '[Jim Watson's book The Double Helix] enabled some activists to mount a campaign in Rosalind's name to improve the lot of women in science. This was no doubt well intentioned and indeed useful, but one side-effect was that Rosalind's male colleagues were to some extent demonized. The most prominent demon seemed to be me. Since then, the Franklin/Wilkins story has often been told as an example of the unjustness of male scientists towards their women colleagues, and questions have been raised over whether credit was distributed fairly when the Nobel Prize was awarded. I have found this situation distressing over the years, and I expect this book is in some way my attempt to respond to these questions, and to tell my side of that story.'
He also writes: 'The later years of my career have been devoted to the exploration of the social issues raised by advances in science. I believe that the tensions in the DNA story may shed some light on how tensions in other spheres might be avoided or addressed.'
Maurice Wilkins died on 5th October 2004.

Contents:
1.The great intelligence debate: science or ideology?
2.Origins
3.The end of IQ
4.First steps to g
5.Second steps to g
6.Extracting g
7.Factor analysis or principaI components analysis
8.One intelligence or many?
9.The Bell Curve: facts, fallacies, and speculations
10.What is g?
11.Are some groups more intelligent than others?
12.Is intelligence inherited?
13.Facts and fallacies

Readership: Anyone interested in measuring intelligence or the debate about what intelligence is

Vast amounts have been written about the nature of intelligence and whether and how it can be measured. The issue has been one of great controversy and polemics have been written on all sides.
The measurement of intelligence has always been closely bound up with statistical ideas. UnfortunateIy, not all of the discussants have had a clear grasp of these statistical concepts. Moreover, for good historical reasons, early work on some of the statistical tools was itself controversial ? though more recent developments have led to deeper understanding which has effectively resolved those statistical controversies. Regrettably not all of the commentators seem aware of these developments. In this book David Bartholomew sets the issues in the context of his seminal work on social measurement. He describes how the main single (latent) dimension of variation in human mental abilities can be measured in terms of manifest test variables, and distinguishes between this and IQ. Within the context of his model he examines the various controversies which have dogged discussions of measuring intelligence.
This book represents a step forward in the debate on measuring intelligence. It is essential reading for anyone concerned with the 'intelligence debate'. It will also make excellent reading for anyone learning about factor analysis, and provides a perfect illustration of the Bartholomew school of measurement models.

Table des matières:
1.Variables aléatoires
2.Espérance mathématique et moments
3.Couples et n-uplets de variables aléatoires
4.Les lois de probabilités usuelles
5.Lois fondamentales de l'échantillonnage
6.Théorie de l'estimation paramétrique ponctuelle
7.Estimation paramétrique par intervalle de confiance
8.Estimation non paramétrique et estimation fonctionelle
9.Tests d'hypothèses paramétriques
10.Tests pour variables catégorielles et tests non paramétriques
11.Régressions linéaire, logistique et non paramétrique

Lecture: Etudiants et professeurs du premier cycle en sciences et sciences appliquées

La théorie des probabilités est introduite sans faire usage de la théorie de la mesure, ce qui fait que le livre est accessible à un public assez large. Les thèmes de la statistique sont classiques: estimation ponctuelle, intervalles de confiance, tests d'hypothèses, régression linéaire. Mais on parle aussi de l'estimation non paramétrique (de la densité, des quantiles, de la fonction de régression) et des tests non paramétriques. La lecture, présuppose une certaine connaissance de mathématiques pour comprendre les démonstrations mais elle reste ouverte à un très grand nombre d'étudiants. Intéressant comme manuel pour l'enseignement est que chaque chapitre se termine par une collection d'exercises, dont les plus difficiles ont un astérisque.

Contents:
1.John Venn and his logic diagram
2.Rings, flags and balls
3.Five and more sets
4.The Gray Code, binomial coefficients and the Revolving-Door algorithm
5.Cosine curves and sine curves
6.Ironing the hypercube
7.Diagrams with rotational symmetry

This book is a delight!
Most school students meet the idea of a Venn diagram to help with probability calculations involving three events: these conveniently involve intersecting congruent circles. When there are more than three events, the question of the Venn diagram construction using closed curves, let alone bounding symmetrical shapes, is much more challenging.
This book describes the development of Venn's ideas from the late nineteenth century to what are recent results. The illustrations are excellent and many are beautiful. The technical mathematical detail is kept to a minimum.
The author, like John Venn before him, is a Fellow of Gonville and Caius College, Cambridge. His interest in Venn' s work and the way it has been developed gives an illuminating account of how research in mathematics is actually done and the excitement of discovering comes across very well. This makes the book ideal for motivation of budding, as well as active, mathematicians and an excellent and attractive addition to bookshelves.
If I were to have a favourite it is the construction of complex Venn Diagrams on a spherical surface to create 'Vennis Balls'.

Contents:
Introduction: Where reason cannot go
1.The energy drain: lmpossible machines
2.The cosmic limit: Unreachable speeds
3.The quantum curtain: Unknowable particles
4.The edge of chaos: Unpredictable systems
5.The circular crypt: Unconstructable figures
6.The chains of reason: Unprovable theorems
7.The computer treadmill: Impossible programs
8.The Big-O bottleneck: Intractable problems

In the words of the author this book "provides a mind bending exploration not into what is doable and knowable ? but what is undoable and unknowable".
As can be seen from the chapter headings the concentration here is on apparently permanent barriers, rather than those which might be overcome in due course as knowledge and technology advance. Of course it is difficult to define the barriers as absolute, but this is an entertaining account for the general reader of the pros and cons as they appear today, together with an account of how our views have evolved.
The author has written several very popular mathematics/science books [including A Mathematical Mystery Tour, Yes We Have No Neutrons, 200% of Nothing] and was Computer Recreations columnist for Scientific American Magazine for eight years.
As a result the style of this book is appropriate for a general readership and it should prove as popular as his other books.

Contents:
Introduction
0.Polytopes and linear programming
1.Matroids and the greedy algorithm
2.Minimum-weight dipaths
3.Matroid intersection
4.Matching
5.Flows and cuts
6.Cutting planes
7.Branch-and-bound
8.Optimizing submodular functions
Readership: Operational researchers, mathematicians, mathematical programmers

The text is designed to be a graduate introduction to combinatorial optimization. It is concerned with the mathematics of the subject; there is only passing mention of applications, nothing about implementation, or sophisticated algorithms that have provable good lower bounds on the number of steps involved. The author, with his Iight but rigorous mathematical writing style, takes delight in reveaIing the stars of combinatorial optimization. The reader does need some mathematical skills, a little knowledge of graph theory concepts, and linear programming. This is an excellent teaching book; I recommend it highly. My regret is that I cannot use it this year as l am not teaching the combinatorial optimization course.

Contents:
1.Uncertainty in forensic science
2.Variation
3.The evaluation of evidence
4.Historical review
5.Bayesian inference
6.Sampling
7.Interpretation
8.Transfer evidence
9.Discrete data
10.Continuous data
11.Multivariate analysis
12.Fibres
13.DNA profiling
14.Bayesian networks

Readership: Forensic scientists, lawyers, teachers and students of statistics seeking interesting applications

Those familiar with the first edition [Short Book Reviews, Vol. 15, p. 41] will notice that the total length has now doubled, as has that of the bibliography. Indeed the number of authors has also doubled, with Franco Taroni joining Colin Aitken, and contributing particularly on the history and philosophy of inference in forensic science, European legal systems, and Bayesian networks.
The doubling in size reflects both coverage of new topics and the explosive growth of the field. There are six new chapters, on Bayesian inference, Sampling, Interpretation, Multivariate Analysis, Fibres and Bayesian Networks. Even now, the book is not exhaustive on applications of statistics to the law, but primarily treats transfer evidence linking a suspected offender with a crime scene. Forensic statistics associated with civil, rather than criminal, cases are not covered, for example regression modelling to establish causation, or discrimination, or to estimate economic Iosses.
The coverage of transfer evidence is, however, wide-ranging and authoritative, including extensive references to the literature. As well as the chapters devoted to DNA profiles and fibres, the evidence types covered include glass, earprints, fingerprints, handwriting, speaker recognition, and paint, among others. A nice feature of the present book is its brief forays into historical context, discussing for example statistical aspects of the infamous C19 French Dreyfus case, the C20 California Collins case that inspired much of the modern academic literature on interpretation of identification evidence. DNA evidence is now a huge field in its own right; the present authors provide a useful introduction, that has been substantially improved and updated since the first edition, but it cannot match the exhaustive treatment of a dedicated book such as that of BuckIeton, Triggs and Walsh (CRC Press, 2004).
The statistical methods discussed are for the most part elementary. There is an introduction to probability, some sampling theory, coverage of some key distributions, and introductions to likelihood and odds ratios, and Bayesian inference. Two more advanced topics have their own dedicated chapters near the end of the book: multivariate analysis and Bayesian networks. The difficult part of the subject is not how to implement any particular statistical technique, but how to formulate appropriate hypotheses and to interpret the statistical results in the context of these. Appropriately, the authors have given considerable attention to these tricky questions of interpretation, including a discussion of common errors and fallacies.

Contents:
1.Introduction
PART I: Extremal Stresses: Extremes, Outliers, Robustness
2.Ordering and extremes: Applications, models, inference
3.Outliers and robustness
PART II: Collecting Environmental Data: Sampling and Monitoring
4.Finite-population sampling
5.Inaccessible and sensitive data
6.Sampling in the wild
PART III: Examining Environmental Effects: Stimulus-Response Relationships
7.Relationship: Regression-type models and method
8.Special relationship models, including quantal response and repeated measures
PART IV: Standards and Regulations
9.Environmental standards
PART V: A Many-Dimensional Environment: Spatial and Temporal Processes
10.Time-series methods
11.Spatial methods for environmental processes

This book brings together a wide range of statistical ideas, concepts and methods relating to environmental research, that formally have been dispersed throughout the literature. The text is well written and the methods are illustrated with interesting examples. Following an introductory chapter describing the kinds of applications considered and the basic concepts and definitions, the author considers the material in five parts. The first is on order-statistics and extreme-value distributions, the identification of outliers, the concept of robust estimation and the accommodation of influential observations. Part two is on sampling methods, beginning with the basic ideas of simple random and stratified sampling applied to finite populations, but also referring to specific methods such as random-set sampling, capture-recapture methods and transect sampling. Part three covers modelling using basic regression analysis, generalized linear models and quantal response models. This is followed by an interesting chapter on the maintenance of environmental standards, set up to control the exposure levels of pollutants or contaminants. The final part brings us right up to date with a description of spatial and temporal models, including time-domain models, frequency-domain models, point and spatial processes, with spatial-temporal models described briefly in the last chapter. The book covers a wealth of statistical concepts as applied to the environment, giving a general overview of the methods developed over many years. One of its main features is that it provides a comprehensive reference source for anyone working on environmental issues.

Contents:
PART I: Causal Inference and Observational Studies
1.An overview of methods for causal inference from observational studies
2.Matching in observational studies
3.Estimating causal effects in nonexperimental studies
4.Medication cost sharing and drug spending in Medicare
5.A comparison of experimental and observational data analysis
6.Fixing broken experiments using the propensity score
7.The propensity score with continuous treatments
8.Causal inference with instrumental variables
9.Principal stratification
PART II: Missing Data Modeling
10.Nonresponse adjustment in government statistical agencies: Constraints, inferential goals, and robustness issues
11.Bridging across changes in classification systems
12.Representing the Census undercount by multiple imputation
13.Statistical disclosure techniques based on multiple imputation
14.Designs producing balanced missing data: examples from the National Assessment of Educational Progress
15.Propensity score estimation with missing data
16.Sensitivity to nonignorability in frequentist inferences
PART III: Statistical Modeling and Computation
17.Statistical modeling and computation
18.Treatment effects in before-after data
19.Multimodality in mixture models and factor models
20.Modeling the covariance and correlation matrix of repeated measures
21.Robit regression: A simple robust alternative to logistic and probit regression
22.Using EM and data augmentation for the competing risks model
23.Mixed effects models and the EM algorithm
24.The sampling/importance resampling algorithm
PART IV: Applied Bayesian Inference
25.Whither applied Bayesian inference?
26.Efficient EM type algorithms for fitting spectral lines in high-energy astrophysics
27.Improved predictions of lynx trappings using a biological model
28.Record linkage using finite mixture models
29.Identifying likely duplicates by record linkage in a survey of prostitutes
30.Applying structural equation models with incomplete data
31.Perceptual scaling
Readership: Academic (Statistics: researchers and practitioners)

This is a collection of articles in a volume dedicated to Professor D. Rubin for his sixtieth birthday. As will be seen from the Contents, the book is divided into four parts, each beginning with an overview of the area. (Incidentally, is it really 'Casual Inference' on pages xiii and 1?)
The issues covered here are missing data, modelling, computation, causal mechanisms, are pervasive in Statistics; also, the application areas are wide-ranging. That said, the 'Rubin Statistical Family Tree' reveals the close connection of most of the authors with Professor Rubin: of the thirty-one chapters, twenty-four have authors who were his students, or students of his students. This gives the book a certain flavour, which will be seen by some as a justifiable unifying theme.
I believe that the book will be a very useful addition to academic libraries. Students (and professors) will be able to look up and learn about particular areas, rather than reading it from cover to cover.

Contents:
1.The Bayesian method
2.Inference and decisions
3.General principles and theory
4.Subjective probability
5.Non-subjective theories
6.Subjective prior distributions
7.Model comparison
8.Robustness and model criticism
9.Computation
10.Markov chain Monte Carlo
11.The linear model
12.Discrete data models
13.Nonparametric models
14.Other standard models
15.Short case studies

Readership: Academic (researchers, practitioners, students); industry and commerce (statistics practitioners)

First, as well as an extra author (Forster now joins the original one, O'Hagan), there is much additional material in this second edition: the old Chapter 7 is now Chapters 7 and 8; the old Chapter 8 is now Chapters 9 and 10; the old Chapter 9 is now Chapter 11, now supplemented by a new chapter, Chapter 12; the old Chapter 10 is now Chapter 14; there are new chapters, Chapter 13 and Chapter 15.
Regarding the philosophy of inference, the authors cannot be accused of sitting on the fence. For example, they say in Section 1.33 that, "It is not surprising that a method [Bayesian] which is fundamentally superior [to the Frequentist Approach], and in particular which makes use of more information, requires more effort to implement". I might argue with the use of "which" here, instead of "that", but I'm sure others would argue more substantially.
As one would expect from these authors the book provides an authoritative, up-to-date, comprehensive ac-count of both Bayesian theory and practicalities. There is much explanation and discussion, which will well serve researchers, practitioners and students. There are plenty of examples and exercises, and the reference list and index are both extensive. Finally, as with the other tomes in the series, the whole is regimented into bite-size, numbered sections for ease of digestion and reference.

Contents:
1.Introduction
2.Variation
3.Uncertainty
4.Likelihood
5.Models
6.Stochastic models
7.Estimation and hypothesis testing
8.Linear regression models
9.Designed experiments
10.Nonlinear regression models
11.Bayesian models
12.Conditional and marginal inference

Readership: Practitioners and researchers in applied and theoretical statistics, from postgraduate level upwards

The concept of an underlying model is fundamental to much statistical thinking today. Here Anthony Davison takes this as the focus for an almost encyclopaedic survey of modern parametric statistics. In my view this mammoth and scholarly undertaking invites comparison with Kendall's original Advanced Theory of Statistics, providing as it does a snapshot of the discipline at the present time. 'Classical' material is presented side by side with recent developments including graphical models, extreme value models, robust estimation and computational techniques. The exposition avoids undue technicality, but indicates the nature of technical requirements where necessary. There are plenty of examples and exercises, many of which will be suitable for teaching purposes. An accompanying library of online sets of data is promised, although as of June 2004 this does not appear to have materialized. Marginal notes provide a mix of historical background information and entertainment. Anybody who is seriously involved in the theory or practice of statistics would be well advised to ensure that they have access to a copy.

Contents:
1.Introduction and notation
2.Data collection designs
3.Kernal equating: Overview, pre-smoothing, and estimation of r and s
4.Kernal equating: Continuization and equating
5.Kernal equating: The SEE and the SEED
6.Kernal equating versus other equating methods
7.The equivalent groups design
8.The single group design
9.The counterbalanced design
10.The NEAT design: Chain equating
11.The NEAT design: Post-stratification equating

APPENDIX A: The delta method
APPENDIX B: Bivariate smoothing
APPENDIX C: Other univariate moments
APPENDIX D: Review of the use of matrices in this book

All three authors work at the Educational Testing Service in Princeton, New Jersey. Their joint opus is a specialized book about research and practice in the field of test equating, which arises from the need to be able to produce tests that can be consistently interpreted over many groups of students who may or may not have answered the exact same questions, at the same or different times.
Chapters 2-6 present the theory; applications follow in Chapters 7-11. The book is nicely laid out, is extremely well written, and is an excellent text for a semester course or a short course. There are sixty-three diagrams and seventy-two references. The book is highly recommended.

Contents:
1.Introduction
2.The characteristic function
3.Linear combinations, products, and ratios
4.Bivariate generalizations and related distributions
5.Mutivariate generalizations and related distributions
6.Probability integrals
7.Probability inequalities
8.Percentage points
9.Sampling distributions
10.Estimation
11.Regression models
12.Applications

Readership: Statisticians interested in continuous multivariate distribution theory, analysis and applications

This monograph is the latest publication to concentrate on a narrow class of very closely related distributions and to examine it in great detail from various aspects. The authors believe that the multivariate t-distributions have been somewhat overshadowed by the multivariate normal, although they provide a more viable alternative for handling real data and are widely used in Bayesian analysis of multivariate data. Their aim was "to collect and present in an organized and user-friendly manner all of the relevant information available in the literature worthy of publication". The first part of the book is mainly devoted to distribution theory; the second is more concerned with estimation and applications. There are about 400 references in the bibliography. The general style and mathematical level are similar to that of Continuous Multivariate Distributions, 1, 2nd edition (2000), S. Kotz, N. Balakrishnan, and N.L. Johnson, Wiley [Short Book Reviews, Vol. 20, p. 41].

Contents:
0.Preliminaries
1.One-sample methods
2.Two-sample methods
3.K-sample methods
4.Paired comparisons and blocked designs
5.Tests for trends and association
6.Multivariate tests
7.Analysis of censored data
8.Nonparametric bootstrap methods
9.Multifactor experiments
10.Smoothing methods and robust model fitting

The author assumes that readers have an introductory statistics course as background knowledge. The text is plainly but attractively set, the writing is nicely conversational and extremely clear. Four computer packages are featured: Resampling Stats, StatXact, S-Plus and MINITAB. The many sets of data throughout vary from real sets of data (attributed, with references), through sets made up or generated but based on experimental ideas (e.g., battery lifetimes), to ones that appear to be simply made up. The balance of these is reasonable. There are a few diagrams and one hundred and three references. The book would be an excellent choice for self-study or for a one-semester course.

Contents:
1.lntroduction
2.Histogram
3.Nonparametric density estimation
4.Nonparametric regression
5.Semiparametric and generalized regression models
6.Single index models
7.Generalized partial linear models
8.Additive models and marginal effects
9.Generalized additive models

Readership: Undergraduate and first-year graduate statistics, mathematics, econometrics and biostatistics students; graduate students, researchers

The book is a downloadable e-book; see www.i-xplore.de. The book is very well written and a pleasure to read with the methods fully illustrated. At the end of each chapter is a valuable summary of the important formulae and methods introduced in the chapter. The book progresses by first motivating, then developing methodology, and finally providing statistical properties. Difficulties are raised at the end of each chapter, and these motivate the development of improvements in the following chapter. There are only a few exercises, and they are typically of a technical nature. The only disappointment is that there is virtually no reference to computing. The illustrations are based on data in references, and it would be very difficult for someone to check the numbers or calibrate their own computations. Discussions on where to find software to implement the methods are needed.

Contents:
1.Introduction
2.Some basic results in probability and statistics
3.Simple linear regression
4.Random vectors and matrix algebra
5.Multiple regression
6.Residuals, diagnostics and transformations
7.Further applications of regression techniques
8.Selection of a regression model
9.Multicollinearity: diagnosis and remedies

Readership: Scientists of aII Ievels needing advanced statistical data analysis and modeling

On page 1, this book is described by "in contrast to other books on this topic [27, 87], we have attempted to provide details of the theory rather than just presenting computational and interpretive aspects." I take exception to the word "just" in this context, because [27] is N.R. Draper and H. Smith's Applied Regression Analysis, 3rd edition, 1998, published by Wiley [Short Book Reviews, Vol. 19, p. 5]. The book [87] is Introduction to Linear Regression Analysis, 3rd edition, 2001, by D. Montgomery, E. Peck and G. Vining.
In the text under review, a lot of mathematical work is added to the regression analyses of data, and the result is a rather uneven presentation which, I think, will please few students. The authors suggest taking two semesters over the material unless some prior knowledge can be assumed.
I was surprised to find several instances of plagiarism of material taken from my own and Harry Smith's book. Parts of their pages 1 and 2 come from our page 45. The data on pages 317 and 319, 320 and 321 come from our pages 314 and 317. Their exercise 7.3 is from our exercise 14C. Their exercise 8.1 is from our exercise 15D. There is no indication at all that their sources are copyrighted.
The list price given inside the front cover is US$195.00 for 436 pages. This per page charge of 45 cents is high, compared with the 16 and 15 cents of the other books mentioned.
This book cannot be recommended.

Contents:
1.Introduction
2.One-way classification
3.Two-way crossed classification without interaction
4.Two-way crossed classification with interaction
5.Three-way and higher crossed classifications
6.Two-way nested classification
7.Three-way and higher nested classifications
8.General balanced random effects model

Readership: Statisticians and experimentalists who use analysis of variance to develop random effects models.

This is Volume I, covering balanced data, of a two-volume work on the analysis of variance used to develop random effects models. The theory and practice of fitting models, where all effects are considered to be random, are discussed in considerable detail for a wide range of experimental situations involving one, two or three factors, including both crossed and nested designs. Following a brief introduction used to elucidate in a unified way the basic results for the random effects analysis of variance, the authors cover, in subsequent chapters, the theory associated with models of increasing complexity, beginning with the simple one-way classification and ending with a description of the general balanced random effects model. Within each chapter for each model, the distribution theory of a variety of classical estimators such as ML, REML, MVU, Stein-type, Naqvi goodness of fit and Hodges-Lehmann-type estimators, is described and these estimators are illustrated with a numerical example supported by computer output derived from SAS, SPSS and BMDP. Bayesian estimators are also considered. The sampling distributions of the estimators of the variance components are covered at length, and confidence intervals and test procedures are developed for the variance components and selected functions of them. This text will provide a useful reference source for theoretical results and practical examples since each chapter is further supported by a comprehensive set of exercises and an extensive reference list.

Contents:
1.Examples of multivariate data
2.Multivariate data and the forward search
3.Data from one multivariate distribution
4.Multivariate transformations to normality
5.Principal components analysis
6.Discriminant analysis
7.Cluster analysis
8.Spatial linear models

Readership: Advanced students of statistics, experimental scientists, statisticians

This book is a companion to Atkinson and Riani (2000, Robust Diagnostic Regression Analysis) [Short Book Reviews, Vol. 21, p. 4]. The idea of the forward search is first of all to identify a subset of the data that is free of outliers; as the search progresses, additional data points are added, and the effect is monitored graphically through appropriate plots, often involving Mahalanobis distances. The objective is to identify outliers, appreciate their influence and if possible discover a data transformation which would result in an overall improvement. For instance, the first and second principal components describe 72.5 per cent of the variance in a transformed data example, compared with only 60.6 per cent for untransformed data. Thus the forward search is an empirical procedure, whose performance needs to be appreciated through applications. The book has many of these, chosen principally from the areas of Chapters 4 to 8, and the Appendix lists in full the seventeen sets of data used. Graphical tools are widely used, resulting in three hundred and ninety figures. Each chapter is followed by extensive exercises and their solutions, and the book could be used as an advanced textbook for multivariate analysis courses. Web-sites provide the relevant software in a range of languages. This book is full of interest for anyone undertaking multivariate analyses, clearly emphasising that uncritical use of standard methods can be misleading.

Contents:
1.Preliminaries
2.Computational geometry
3.Neighborhood graphs
4.Class cover catch diagrams
5.Cluster catch diagrams
6.Computational methods

Readership: Advanced undergraduate and graduate students of statistical pattern recognition

As far as I am aware, this is the first book to bring together the two topics of random graphs and statistical pattern recognition. It is not about graphical models or belief networks, a familiar modern intersection of statistics and graph theory, but is about the use of the random proximity and neighborhood graphs which arise in statistical pattern recognition problems. The aim of the authors is that the merger will enhance both communities. Speaking from the perspective of a researcher in pattern recognition, I think that the authors have succeeded. The ideas of random graphs in pattern recognition have always struck me as elegant, and it is nice to see them brought together in such a clear way.
The opening chapter provides background material on the requisite concepts of graph theory and statistical pattern recognition, both supervised and unsupervised. Nearest-neighbour ideas are not only an old staple of statistical pattern recognition, but they also have elegant mathematicaI properties and powerful classification properties. Not surprisingly, then, they figure prominently in this book. Another attractive feature is the wide range of different pattern recognition problems and applications used to illustrate the ideas.
For anyone who wishes to gain a sound understanding of the ideas of pattern recognition in general, as well, of course, as anyone researching in the area of random graphs applied in pattern recognition, this book would be well-worth reading. It is clearly and accessibly written, and nicely conveys the power, breadth and applicability of some very elegant ideas.

Contents:
PART I: Preliminaries
1.Introduction
2.General principles in random variate generation
3.General principles for discrete distributions
PART II: Continuous Univariate Distributions
4.Transformed density rejection
5.Strip methods
6.Methods based on general inequalities
7.Numerical inversion
8.Comparison and general considerations
9.Distributions where the density is not known explicitly
PART III: Discrete Univariate Distributions
10.Discrete distributions
PART IV: Random Vectors
11.Multivariate distributions
PART V: Implicit Modeling
12.Combination of generation and modeling
13.Time series (Authors M. Hauser and W. Hörmann)
14.Markov chain Monte Carlo methods
15.Some simulation examples

Readership: All users of simulation, operational research workers, statisticians

Many statistical methods make use of simulation. We now understand how to simulate standard univariate and multivariate random variables, but what do we do when we encounter a non-standard situation? Suppose for instance we want to simulate from a general stable law, specified in terms of its characteristic function, or from a random variable specified in terms of its probability generating function ? how do we proceed? What if we want to simulate from a t-distribution with parameter 2.3? What if the distribution is specified by means of its hazard function? This fascinating book is the place to look for answers to questions such as these. It brings together many of the recent developments in the area, and makes them available through the C library written by the authors, and called UNU.RAN. The authors clearly describe all the basic methods, but where the book is particularly interesting is when it presents a variety of "automatic" methods. For instance, the inversion method can be applied to a wide range of distributions by making use of numerical integration. Chapter 8 reviews the automatic methods that have been covered for continuous random variables, and makes recommendations so that users can decide which method is best for their particular application. Chapter 10 does the same for discrete distributions. The book ends with details of simulating time-series, MCMC methods, and applications in finance. This book is essential reading for users of simulation, and is destined to become a classical reference for the area.

Contents:
1.Introduction to biostatistics
2.Biostatistical design of medical studies
3.Descriptive statistics
4.Statistical Inference: Populations and samples
5.One- and two-sample inference
6.Counting data
7.Categorical data: Contingency tables
8.Nonparametric, distribution-free and permutation models: Robust procedures
9.Association and prediction: Linear models with one predictor variable
10.Analysis of variance
11.Association and prediction: Multiple regression analysis and linear models with multiple predictor variables
12.Multiple comparisons
13.Discrimination and classification
14.Principal component analysis and factor analysis
15.Rates and proportions
16.Analysis of time to an event: Survival analysis
17.Sample size for observational studies
18.Longitudinal data analysis
19.Randomized clinical trials
20.Personal postscript

Readership: Health professionals, introductory biostatistics students, lecturers

This new edition of a volume first published in 1993 [Short Book Reviews, Vol. 13, p. 37] is to be welcomed. The original authors have been augmented by two "experts" in "all things modern and statistical" (provide your own music). Two new chapters, entitled "Longitudinal Data Analysis" and "Randomized Clinical Trials", have been added. The latter was, surprisingly perhaps, absent from the first edition but the former clearly allows discussion of important recent developments such as generalized estimating equations, mixed models and approaches to missing data. While much of the rest of the book is little changed, this is appropriate and the presence of some marked changes indicates the thought that has been given to the new edition. The authors have also created a set of Web appendices for suitable material. Although a large volume, the first edition of this book was unusual in attempting to convey the original authors' personal pleasure in their subject. This intent is still evident. While, as for the first edition, questions might be raised about the ordering of topics and even some specific recommendations, this updated edition will help to ensure the ongoing usefulness of this valuable resource.

Contents:
1.Introduction
2.Basic concepts from traditional statistical analysis
3.An overview of the Bayesian approach
4.Comparison of alternative approaches to inference
5.Prior distributions
6.Randomised controlled trials
7.Observational studies
8.Evidence synthesis
9.Cost-effectiveness, policy-making and regulation
10.Conclusions and implications for future research

Readership: Medical statisticians, healthcare providers, healthcare policy-makers

This important book presents the case for the use of Bayesian statistical methods in medical and clinical applications. It takes many familiar areas of clinical assessment (experimental and observational studies, sequential trials) and argues for the use of the probabilistically coherent Bayesian approach in the assessment of efficacy, preference and evidence. It contains interesting discussions on an "integrated approach" to health care provision, and persuasively argues that the Bayesian framework is the natural one within which policy decisions and regulations should be made. The technical material is presented in an accessible style, and the examples given clearly illustrate the principles under discussion. I think this book is an essential read for those interested in an evidence-based approach to medicine, and for decision-makers in healthcare.

Contents:
1.Disease mapping basics
2.Bayesian hierarchical modelling
3.Multilevel modelling
4.WinBUGS basics
5.MLwiN basics
6.Relative risk estimation
7.Focused clustering: The analysis of putative health hazards
8.Ecological analysis
9.Spatially-correlated survival models
10.Epilogue

APPENDIX 1: WinBUGS code for focused clustering models
APPENDIX 2: S-Plus functions for conversion of GeoBUGS format

Readership: Researchers in disease mapping epidemiology, graduate students in statistics

This useful book outlines the models used in statistical disease mapping, and gives details of how the models can be implemented using two packages: WinBUGS and MLwiN. Bayesian analysis is the main focus, but some pure likelihood methods are also discussed. The book makes four contributions: it provides an introduction to the models used, discusses basic Markov chain Monte Carlo (MCMC) procedures, introduces the necessary software, and illustrates the use of the software on familiar examples. I think the most important contribution is the last; it is vitally important, if Bayesian methods are to be used routinely by health scientists, that the methods are seen to be routinely implementable on widely available platforms. MCMC methods are often viewed by non-statisticians as being rather difficult to formulate and time consuming to implement. This book addresses and redresses that impression in the field of spatial epidemiology.

Contents:
1.Introduction
2.Measures of accuracy for binary tests
3.Comparing binary tests and regression analysis
4.The receiver operating characteristic curve
5.Estimating the ROC curve
6.Covariate effects on continuous and ordinal tests
7.Incomplete data and imperfect reference tests
8.Study design and hypothesis testing
9.More topics and conclusions

Most medical statisticians will see data that relate to a diagnostic test at some time in their career. Concepts such as sensitivity, specificity and predictive values are well recognized by many but there is often less understanding of how methodology used in this area relates to more general developments in statistical research. Also the literature on diagnostic tests has primarily been in subject- matter journals with little emphasis being given to inferential details. This book now provides a welcome overview of diagnostic testing and indeed, as reflected in the title, of classification and prediction issues more generally.
Beginning with the simplest concepts and binary tests, the book goes on to outline methods for more complicated situations. Particular attention is given to covariate effects and the comparison of tests. Later chapters deal with imperfect data sources, study design and an overview of additional topics, some of which may motivate future research efforts. It is a well-written book and provides both a clear description of methods and a serious discussion of their statistical properties. Each chapter ends with some overview remarks and exercises. The preface indicates that the book is aimed at 'the practicing statistician' but with some sections directed to the 'academic research biostatistician'. I believe both, and particularly those who wear two hats, will find the book of considerable value.

Contents:
1.Population proportion of prevalence
2.Risk difference
3.Relative difference
4.Relative risk
5.Odds ratio
6.Generalized odds ratio
7.Attributable risk
8.Number needed to treat

This volume covers the main measures of risk used by epidemiologists and clinical trialists. Each is dealt with systematically, in similarly structured chapters which describe and compare the appropriate point and interval estimators to be used under alternative sampling designs. The content is very detailed, supported by extensive and up-to-date references, many from the author himself who is undoubtedly an expert in the field. The style is extremely technical, making this textbook a very useful reference for biostatisticians working in epidemiology or clinical trials but not, in my view, an accessible source for applied researchers.
The chapters are independent and thus each is itself a separate reference source. Every chapter begins with a brief and clear introduction to the measure being covered, namely risk (or prevalence), risk difference, relative risk (i.e. risk ratio for prevalent data and rate ratio for incidence data), and relative difference (i.e. relative risk reduction), odds ratio, generalized odds ratio, attributable risk and numbers needed to treat (i.e. the reciprocal of the risk difference). The introduction is followed by the point estimators and exact and asymptotic interval estimators appropriate for different designs. These include inverse sampling which is shown to be very useful when standard binomial sampling would lead to biased estimates. In all chapters the designs discussed include those obtained after stratification by potential confounders, clustering and paired-sampling. Several examples of real data are used to compare and discuss the different estimators. Specific references support every result and several exercises are given at the end of each chapter. Thus a very broad overview of the most commonly used measures of risk and of the many available interval estimators is given by the author. I was surprised, however, not to find a more prominent theoretical discussion of how the different asymptotic interval estimators relate to each other through their various approximations to the likelihood function: the appendix indirectly covers this but only from a technical point of view.
At first sight the amount of details and references may appear overwhelming. However, these are systematically organized so that, despite this being a very technicaI book, it is an excellent reference. More importantly, this book is likely to have an impact for further studies because of its demonstration of the advantages of inverse sampling designs when events are rare and resources limited.

Contents:
PART I: Overview
1. Introduction
2. Vectors and matrices
3. Regression analysis
PART II: Analysis of Interdependence
4. Principal components analysis
5. Exploratory factor analysis
6. Confirmatory factor analysis
7. Multidimensional scaling
8. Cluster analysis
PART III: Analysis of Dependence
9. Canonical correlation
10. Structural equation models with latent variables
11. Analysis of variance
12. Discriminant analysis
13. Logit choice models

Readership: Application orientated researchers teaching graduate courses in statistics to students in fields of application

This book is not merely an update of an earlier version but a rebirth giving a fresh look at applying multivariate analysis in this day and age. Although application-orientated, it avoids a "black box" approach but relies on an intelligent, critical intuitive grasp of the working of techniques. Therefore each chapter starts with an intuitive explanation based on the underlying geometrical properties of the algebra. In the mechanics section some more mathematical derivations are provided in modular form. Every technique is further explained by a practical problem and questions relating to the application of the method. Each chapter is concluded by a learning summary, selected readings and exercises.
The geometrical implications of vector and matrix operations are reviewed as well as multiple linear regression. This lays the foundation for more sophisticated multivariate techniques. Part II deals with analyzing interdependence. A sound introduction is provided on how and when principal components, exploratory factor analysis and confirmatory factor analysis should be applied. The distinction made between classical metric multi-dimensional scaling (MDS), nonmetric MDS, individual difference scaling and MDS of preference data illustrates the wide area of potential applications of MDS. A book of this scope cannot provide a comprehensive account of modern clustering techniques, but the fundamental ideas behind agglomerative clustering and partitioning are clearly conveyed; linkage methods, Ward's method, k-th nearest neighbour and k-means are introduced.
Part III is devoted to the analysis of dependence. The traditional topics of canonical correlation and multiple analysis of variance (including repeated measures designs and controlling for covariates) are covered together with more modern topics like structural equations modelling and latent variables. The fundamentals of classical two-group as well as multiple group discriminant analysis are provided, enlightened by several diagrams. Binary as well as multinomial logit models are introduced and their relationship with discriminant analysis explored.
Together with the book, a CD is provided with MDS programs Kyst, Mdpref and Sindscal as well as numerous sets of data from many fields of application. The book is software independent and the sets of data used in the example problems and for exercises are provided for Excel, MINITAB, SAS, S-Plus, SPSS and Stata as well as in ASCII format.
This well-written and well-illustrated book is a real asset on the desk of every practicing statistician.

Contents:
An ancient Pythagorean view of today's statistics
1. Introduction and outline of the book
2. Elementary statistics
PART I: Sampling Theory and Autocorrelations
3. Alternative approach to unequal probability sampling
4. A general rho theory of survey sampIing
5. Variance estimation for some standard designs
6. Multistage and cluster (sub)sampling
7. Systematic sampling
PART II: Variance Estimation in Complex Surveys
8. Estimation of the sampling autocorrelation ñz
9. Variance approximations
10. A simulation study
PART III: Minimum Variance Estimators
11. The regression estimator revisited
12. General restriction estimator in multisurvey sampling
13. Weighting procedures

Parts I and II of this book on sample survey theory present a novel approach to expressing the well known HorvitzThompson estimator of a finite population total, its variance and corresponding variance estimators. The estimator is expressed as a simple mean and its variance in terms of a sampling auto-correlation coefficient (rho) for general sampling designs, similar to the wellknown representation of the estimator in singlestage cluster sampling in terms of an intra-cluster correlation coefficient. The rhocoefficient, however, involves the joint inclusion probabilities. Therefore, approximate methods of estimating rho and the variance are studied in the context of randomized systematic sampling with unequal probabilities (also called dollar unit sampling). The author has not studied the wellknown Rao-Sampford method of unequal probability sampling that Ieads to a nonnegative unbiased variance estimator. This variance estimator can be calculated easily using the SAS sampling package for any sample size. Exact variance estimators should be preferred unless one can demonstrate that the approximate versions are more stable and lead to small relative bias. Part III uses ideas from the econometrics literature to provide efficient estimators using information contained in linear or nonlinear restrictions on the parameters. The proposed approach can handle a variety of estimation problems and it leads to the well-known optimal linear regression estimator when the population means of auxiliary variables are known. The author provides Pythagorean perspectives to explain the proposed methods.

Contents:
1. Descriptive methods for categorical data
2. Descriptive methods for continuous data
3. Probability and probability models
4. Estimation of parameters
5. Introduction to statistical tests of significance
6. Comparison of population proportions
7. Comparison of population means
8. Correlation and regression
9. Logistic regression
10. Methods for count data
11. Analysis of survival data and data from matched studies
12. Study designs

Readership: Professionals and beginning graduate students in public health, dentistry, nursing, medicine, biomedical sciences

The emphasis in this book is on statistical methods relevant to health studies. For example, it starts with a chapter on counts, proportions, ratios and rates (likely to motivate readers); the section on probability is written mainly in terms of screening tests, confidence intervals for odds ratios and the Mantel-Haenszel method are covered. Chapters 9 and 11 are clearly on methods used in the analysis of health and medical data, and so is Chapter 10 which covers the Poisson regression model. The chapter on study designs deals with sample size determination.
The book is clearly written. Some chapters end with brief notes on the fundamentals which flesh out the theory. Samples of SAS procedures are given throughout, and there are Excel instructions in the earlier chapters.
My major criticism relates to the data used in examples and exercises. Some are real and their source is given, others appear to be real but do not give a source, and the remainder could well be fictional. Few of the studies cited are recent but were done in the 1970s or 1980s. Some of the longer sets of data are available from the author, but the book would be of more value were they included on a disk or at Ieast available on a web page.

Contents:
1.Introduction
2.Diagnostic checks for univariate linear models
3.The multivariate linear case
4.Robust modeling and diagnostic checking
5.Nonlinear models
6.Conditional heteroscedasticity models
7.Fractionally differenced process
8.Miscellaneous models and topics

There have been several excellent monographs on the diagnostics of linear models, but this is the first and possibly definitive one for stationary time series modeling. It is of great value in bringing together the diverse literature on the topic, over three hundred references are given, and integrating them into a coherent whole. Although not over-faced with formulae, and predominantly informative discussion and many illustrations, main results are stated clearly as lemmas and theorems. I hope this will not put off applied workers to whom this work is properly directed. The main theme of the monograph is the development and application of the generic autocorrelation-based portmanteau goodness-of-fit test to many different models, this being seen as the 'chi-square test' of time series. Whatever type of time series model you are fitting, linear or nonlinear, volatile or not, turn to this monograph for help in testing its goodness-of-fit.

Contents:
1.Introduction
2.Probability and statistical models
3.Returns
4.Time series models
5.Portfolio theory
6.Regression
7.The capital asset pricing model
8.Option pricing
9.Fixed income securities
10.Resampling
11.Value-at-risk
12.GARCH-models
13.Nonparametric regression and splines
14.Behavioural finance

Readership: Undergraduate or master students in engineering, mathematics, statistics and economics

The inherent interaction of statistical and financial modelling makes this book a very useful and motivating instrument with which to introduce students from engineering, mathematics, statistics and economics to study statistics and/or finance. Despite being written in a very accessible style which avoids too technical details, the manuscript succeeds in covering relatively recent topics from statistics and finance, like the bootstrap, penalized splines, some VaR estimation models and behavioural finance. Several financial applications of the introduced statistical methods are presented, including for instance the fitting of volatility smiles with polynomial regression, the estimation of a continuous forward curve, the incorporation of estimation risk in portfolio choice with bootstrap methods and the estimation of a tail index in the context of risk management. Students having gained confidence with the material of this book can also be expected to be ready for advanced topics not covered in the manuscript, like for instance generalized method of moments statistics or indirect inference methods.

Contents:
1.Options
2.Option valuation preliminaries
3.Random variables
4.Computer simulation
5.Asset price movement
6.Asset price model: Part I
7.Asset price model: Part II
8.Black-Scholes PDE and formulas
9.More on hedging
10.The Greeks
11.More on the Black-Scholes formulas
12.Risk neutrality
13.Solving a nonlinear equation
14.Implied volatility
15.Monte Carlo method
16.Binomial method
17.Cash-or-nothing options
18.American options
19.Exotic options
20.Historical volatility
21.Monte Carlo Part II: Variance reduction by antithetic variates
22.Monte Carlo Part III: Variance reduction by control variates
23.Finite difference methods
24.Finite difference methods for the Black-Scholes PDE

This is an introductory level text, to the extent that the author claims that no background in probability, statistics, or numerical analysis is needed, though it does require 'a working knowledge of first year calculus'. It is pitched at the same sort of level as the classic text by Wilmott, Howison, and Dewynne (The Mathematics of Financial Derivatives, CUP, 1995), though it is narrower in scope, with less emphasis on PDEs and more on stochastic modelling and simulation. The author summarizes the key features of the book as being: (i) detailed derivation and discussion of the basic log-normal asset price model; (ii) roughly equal weight given to binomial, finite difference and Monte Carlo methods; (iii) heavy use of computational examples and figures; (iv) stand alone MATLAB code implementations of the main algorithms. It gives equal weight to applied mathematics, stochastics, and computational algorithms, and is structured so that each chapter could be taught in one hour, making it convenient for teaching. It would be worth considering as a text for students new to the area, but who were not intending to specialize in the topic. It could serve as a good motivating medium through which to introduce the various statistical and mathematical tools which it uses.

Contents:
PART I: Finance and Risk Management
1.Potpourri
2.Making economic decisions under uncertainty
3.Expected utility
4.Probability finance
5.Derivatives finance
PART II: Mathematical and Computational Finance
6.Options and derivatives finance mathematics
7.Options and practice
8.Fixed income, bonds and interest rates
9.Incomplete markets and stochastic volatility
10.Value at risk and risk management

Readership: Students and practitioners
interested in finance, economics,
risk management

The author claims that this is not just another book on mathematical finance. It indeed contains several discussions on economic issues which a more mathematically oriented text would typically not include. The style is rather informal, however, for the more technical bits, the reader will have to sharpen his mathematical pencil somewhat. It is not clear to me how the straddle between the more and less formal discussions will please the intended readership; time will tell. Towards the end, I got the feeling that the author was trying hard to put all relevant risk management issues on the table; needless to say that for several concepts, this could only be achieved at the cost of lack of depth. All in all, this book gives a refreshing approach.

Contents:
1.Financial modelling beyond Brownian motion
2.Basic tools
3.Lévy processes: Definitions and properties
4.Building Lévy processes
5.Multidimensional models with jumps
6.Simulating Lévy processes
7.Modelling financial time series with Lévy processes
8.Stochastic calculus for jump processes
9.Measure transformations for Lévy processes
10.Pricing and hedging in incomplete markets
11.Risk-neutral modelling with exponential Lévy processes
12.Integro-differential equation and numericaI methods
13.Inverse problems and model calibration
14.Time inhomogenous jump processes
15.Stochastic volatility models with jumps

Readership: Graduate students and researchers in financial mathematics as well as mathematicians working in banks and financial institutions

This book is an extremely rich source of information for recent developments in the use of jump processes in financial modelling, in particular the use of Lévy processes. The contents list speaks for itself in this respect. The book as a whole is non-assuming in the sense that the mathematical and financial background the reader would seem to need is little more than what one would expect from a masters level education from any good European mathematics department offering courses in financial stochastics. The authors work at a comfortable mathematical pace choosing carefully which proofs to include and exclude and never losing sight of financial interpretation and application. The book comes with many additional perks. For example, many examples and local summaries, a balanced perspective on the shortfalls of the theory being presented, making the effort to show how standard results and expressions for semi-martingales look like for Lévy processes and clear referencing for further reading. The book is also spiced with some very interesting historical notes about prominent (French) mathematicians whose work has ultimately contributed to the foundations of financial stochastics.
The authors conclude the main body of their text by saying: "We hope that the present volume will encourage more researchers and practitioners to contribute to this topic and improve on our understanding of theoretical, numerical and practical issues related to financial modelling with jump processes". I am quite convinced that this goal will be achieved.

Contents:
Acronyms and Abbreviations
Contributors
200 articles, alphabetically by subject
Author Index
Subject Index

This is the third volume in the Wiley Series in Biostatistics, and is based on the existing Encyclopedia of Biostatistics, with articles updated and modified where necessary. In addition 42 new articles have been added. Invariably opinions on the choice of topics will differ, but I found that this book in general gave a good coverage of the area. I also thought that the articles were commendably brief and informative. It is impossible to give much detail in the constrained space available, but the articles give a good overview and sufficient ideas for further reading, so that this is an excellent first reference. I expect to use this book frequently in the future and will be very glad to have it on my bookshelves.

Contents:
1. Introduction
2. The impact of mismeasured continuous variables
3. The impact of mismeasured categorical variables
4. Adjusting for mismeasured continuous variables
5. Adjusting for mismeasured categorical variables
6. Further topics

Readership: Graduate students, practitioners, and researchers in statistics and biostatistics, and epidemiologists with particular interest in quantitative methods

The title of this book implies a breadth greater than the content. In fact, it is only concerned with models for predicting some outcome variable from explanatory variables in the case when an explanatory variable is mismeasured. The author describes the book as neither a textbook nor a research monograph, but rather providing a mix of expository and research-oriented material, and I think this is an accurate description. Its integration of material on continuous and discrete situations is fairly novel. The book adopts the Bayesian perspective, and groups much of the mathematical development at the end of each chapter. It provides a useful introductory overview of the, now large, research literature, and I would recommend it to anyone new to the area.

Contents:
1. Introduction
2. The binary model
3. Maximum likelihood estimation of the binary logit model
4. Some statistical tests and measures of fit
5. Outliers, misclassification of outcomes, and omitted variables
6. Analyses of separate samples
7. The standard multinomial logit model
8. Discrete choice or random utility models
9. The origins and development of the logit model

This slim volume is designed for newcomers to the subject and I suggest that one initially reads it in a cursory manner before returning to any topic in depth.
The text provides an easy-to-read introduction to the theory underlying logit analysis and gives a thorough exposition of the technique of estimation; with computational routines for the logit model forming a common part of general statistical packages. The author makes repeated use of a set of data on private car-ownership of Dutch households which can be downloaded from the Cambridge University Press website. These data form the basis of the worked examples which will be invaluable for those new to logit analysis.
The author has assumed that the reader is familiar with ordinary linear regression and with the associated estimation theory and matrix algebra. An historical perspective on the origins and development of the logit model is provided in the final chapter.

Contents:
1. Introduction
2. Reliability for continuous scale measurements
3. Measures of 2 x 2 association and agreement of cross classified data
4. Coefficients of agreement for multiple raters and multiple categories
5. Assessing agreement from dependent data
6. Sample size requirements for the design of a reliability study
7. Workshops

The wealth of literature that addresses the issues of interobserver reliability and agreement can be overwhelming for the researcher. There is confusion surrounding terminology, and to date there has been little attempt to produce a coherent summary of current techniques. This book is much needed, and successfully summarizes some of the more common techniques in current use, identifying appropriate measures and modelling approaches for a number of different situations. The reader is provided with clearly worked examples and workshops that are likely to be very helpful for those undertaking their own analysis. Of particular use to the researcher is the chapter on samplesize requirements as this issue is important and frequently not addressed.
A reasonably high level of statistical understanding is assumed, and the detail provided might at times be confusing for the non-statistician. Both statisticians and non-statisticians would benefit from further clarification of the terminology used and perhaps more detailed discussions on the interpretation of some of the measures presented.

Contents
1. Introduction
2. Nonparametric methods
3. Parametric models
4. Regression models
5. The Cox proportional hazards model
6. Model checking: Data diagnostics
7. Additional topics
8. Censored regression quantiles

This book grew out of a short course on survival analysis, and retains features of the latter, including concise coverage of basic topics, a point form summary of objectives at the start of each chapter, and instructions (with code) on how to implement the methods discussed using S or R software. The authors express their admiration for D.G. Kleinbaum's 1995 Springer book Survival Analysis: A Self-Learning Text and adopt several of its features. The result is a well-written, accessible introductjon to basic survival analysis methods. The book does not offer much theoretical background and its choice and coverage of topics is limited; therefore it seems best suited for short courses, self-study, or as supplementary material on S implementation. One interesting feature deserves special mention: a final chapter on nonparametric regression quantile methods written by Stephen Portnoy. This useful methodology is not covered in other books on survival analysis.

Contents:
1. Randomization and the clinical trial
2. Issues in the design of clinical trials
3. Randomization for balancing treatment assignments
4. Balancing on known covariates
5. The effects of unobserved covariates
6. Selection bias
7. Randomization as a basis for inference
8. Inference for stratified, blocked, and covariate-adjusted analyses
9. Randomization in practice.
10. Response-adaptive randomization
11. Inference for response-adaptive randomization
12. Response-adaptive randomization in practice
13. Some useful results in large sample theory
14. Large sample inference for complete and restricted randomization
15. Large sample inference for response-adaptive randomization

Readership. Graduate students in biostatistics, practising statisticians and clinical trialists

This is a scholarly review of the whole topic of randomiziation in clinical trials. The full range of available methods is described, from simple randomization through the biased-coin design and urn models to response-adaptive schemes that attempt to randomize more patients to the better treatment. Their abilities to achieve desired allocation ratios, while minimizing accidental bias and predictable allocations, are explored. Sufficient mathematical detail is given for the interested reader to understand the derivations, but formal proofs are avoided. The authors are proponents of randomization-based inference, which is systematicaIly described, and model-based inference is also covered. Alongside the theoretical development is discussion of practical considerations and examples.
With its extensive sets of problems and discussion exercises, the book should be very useful in graduate courses in biostatistics. Researchers will find both a valuable guide to the literature and a number of suggestions for future research. The statistician working in clinical trials will be mainly interested in reading the variety of available randomization schemes, but may be disappointed by the lack of concrete recommendations. For all of these groups of statisticians, this book will form a valuable reference work.

Contents:
1. Survival analysis
2. Some non-parametric procedures
3. Modelling survival data
4. Model checking in the Cox regression model
5. Parametric proportional hazards models
6. Accelerated failure time and other parametric models
7. Model checking in parametric models
8. Time dependent variables
9. Interval censored survival data
10. Sample size requirements for a survival study
11. Some additional topics
12. Computer software for survival analysis

Readership: Statisticians in any field; students of statistics, researchers doing their own analyses

This book is a revised version of the author's original book, which was published in 1994 [Short Book Reviews, Vol. 14, p. 25]. The content is reorganized and extended to include developments that have occurred over the intervening years. The chapter headings give an idea of the scope of the book. The counting processes approach to survival analysis is only mentioned incidentally, but references are given.
In the first edition, survival analysis programs offered by a number of statistical packages, were reviewed. In this edition the author has restricted himself to detailed examples of the SAS survival analyses procedures. Comments on their outputs are given with mention of parts, such as "Type III analysis of effects", that should be ignored!
The title of this book may suggest that it would be only useful in medically related research. This is far from the case. It could profitably be used by anyone, who has survival data to analyze. A complete novice, armed with a SAS output, who carefully follows the examples of model fitting and checking in chapters three and four, should not go far wrong! However the scope of the book goes far beyond that. Throughout it aims to convey an appreciation of the theoretical underpinnings of the subject. Of particular value are chapters five and six on the parametric proportional hazards and the accelerated failure time models, respectively. The author shows how each model is produced by different parameterizations of the Weibull and exponential distributions. As in all other chapters, these models are fitted to an appropriate set of data and the parameters interpreted in the context of the data.
For a student following a course in survival analysis or the applied statistician, this book is full of practical insights, and exemplifies the very best in statistical practice. The author has a rare gift for lucid exposition. The book is highly recommended.

Contents:
1. Introduction
2. Functions of survival time
3. Examples of survival data analysis
4. Nonparametric methods of estimating survival functions
5. Nonparametric methods for comparing survival distributions
6. Some well-known parametric survival distributions and their applications
7. Estimation procedures for parametric survival distributions without covariates
8. Graphical methods for survival distribution fitting
9. Tests of goodness of fit and distribution selection
10. Parametric methods for comparing two survival distributions
11. Parametric methods for regression model fitting and identification of prognostic factors
12. Identification of prognostic factors related to survival time: Cox proportional hazards model
13. Identification of prognostic factors related to survival time: Nonproportional hazards models
14. Identification of risk factors related to dichotomous and polychotomous outcomes

Readership: Biostatisticians, graduate students in biostatistics and epidemiology

This is the third edition of a well-known monograph [Original 1980; Review of second edition Short Book Reviews, Vol. 13, p. 3]. New material includes mathematical details related to likelihood inference, a broader overview of models with nonproportional hazards and an expanded chapter on linear logistic regression. The result is a comprehensive and clearly written textbook which is also accessible to applied researchers with limited mathematical background. The examples are very helpful and are often shown with the relevant programming codes for BMDP, SAS and SPSS.
The first part of the book, aimed at introducing the reader to the special issues arising with survival data, includes a chapter of examples of published data. This has the notable advantage of giving an immediate feel for the complexities of dealing with survival data. It has, however, the disadvantage of using terminology and methods still unfamiliar to a newcomer to the subject. Some simple description of the reasons why certain methods are used (or in some cases should not be used) would make this chapter more useful.
Part II of the book introduces nonparametric methods for survival analysis. Unlike other textbooks in survival analysis it includes references to relative survival methods and standardized mortality and incidence ratios, thus making the connection with related material which is usually confined to textbooks in descriptive epidemiology.
Part III deals with parametric methods and is the more extensive of the book, covering nine of its fourteen chapters. The contents are well organized and detailed, always supported by real examples of data and often by software codes. Of interest are the sections on competing risks and the recurrent events models.
Overall this is a very good reference and an excellent, if somewhat overdetailed, textbook. Throughout there are interesting exercises, some of which may be useful for drafting examination questions as well as for illustrations.

Contents:
1. Introduction
2. Parametric regression
3. Scatterplot smoothing
4. Mixed models
5. Automatic scatterplot smoothing
6. Inference
7. Simple semiparametric models
8. Additive models
9. Semiparametric mixed models
10. Generalized parametric regression
11. Generalized additive models
12. Interaction models
13. Bivariate smoothing
14. Variance function estimation
15. Measurement error
16. Bayesian semiparametric regression
17. Spatially adaptive smoothing
18. Analyses
19. Epilogue

Readership: Academic (researchers and postgraduate students in Statistics, Economics, Finance); Users of Statistics (Industry, Medical Research, ...)

In their preface, the authors say that the book is suitable for several audiences. These include those with 'only a moderate background in regression', those 'who have a good working knowledge of linear models', and 'experts on smoothing'. This might seem to be a little ambitious, but there is a lot of material here and it is very sympathetically presented. As the authors say, this is a user-friendly book, with lots of graphs and pictures, and examples and case studies from a variety of fields.
Most chapters have Bibliographical Notes at the end, and the earlier ones also have a Summary of Formulae. Computation is dealt with in Appendix B, where the matrix formulae are given first and then S-Plus and Matlab code. Some software is also referred to, including S-Plus functions and SAS procedures.
I would recommend this book to anyone interested in the field. It is very readable, informative without being heavy, and (excellent news) comes in a paperback version as well as hardback.

Contents:
1. Procedures for a single sample
2. Procedures for two independent samples
3. Procedures for paired samples
4. Procedures for several independent samples
5. Analysis of block designs
6. Independence, correlation, and regression
7. Computer-intensive methods

The book covers the classical nonparametric statistical methods both for complete samples and for randomly right censored samples. The approach is intuitive rather than mathematical. Each chapter has an Appendix A with some mathematical derivations and an Appendix B with computer programs in the SAS language. This unique format plus the fact that each chapter also has a number of problems makes the book also interesting for teaching purposes.

Contents:
1. Introduction
2. An overview
3. Minimax distinguishability
4. Sharp asymptotics I
5. Sharp asymptotics II
6. Gaussian asymptotics for power and Besov norms
7. Adaptation for power and Besov norms
8. High-dimensional signal detection

Readership: Mathematical statisticians with interest in nonparametric statistical inference

The book deals with nonparametric goodness-of-fit testing problems from the literature of the past twenty years. The setting is based on the asymptotic variant of the minimax approach. The key element is the construction of asymptotically least favourable priors for a wide class of nonparametric testing problems. The method leads to various types of asymptotically optimal tests. The problems are studied within Gaussian models. It is a theoretical book with mathematical results rather than solutions to applied problems in engineering or medicine. The proofs of the theorems are very detailed and many details are in the appendix of more than one hundred pages.

Contents:
1. Introduction and fundamentals
2. Order statistics, quantiles, and coverages
3. Tests of randomness
4. Tests of goodness-of-fit
5. One-sample and paired-sample procedures
6. The general two-sample problem
7. Linear rank statistics and the general two-sample problem
8. Linear rank tests tor the location problem
9. Linear rank tests for the scale problem
10. Tests of the equality of k independent samples
11. Measures of association for bivariate examples
12. Measures of association in multiple classifications
13. Asymptotic relative efficiency
14. Analysis of count data

Readership: Statisticians, final year undergraduate and graduate statistics students

The facts that the first edition of this book was published in 1971 and that it is now in its fourth and revised edition are testimony to the book's success over a long period. The Iast revision was in 1992. The authors' goals in this latest edition were to bring the material covered into the twenty-first century and to make the material more user friendly.
New material and references have been added, and some material has been reorganized. Greater emphasis has been placed on P values. The computer packages for which sample output of problem solutions is included are now MlNITAB, SAS, STATXACT, and SPSS, reflecting current teaching and professional usage. Exact solutions obtained by hand are also given.
Apart from the first chapter, every chapter starts with an introduction and finishes with a summary and problems. There are both theoretical and data-based problems, and answers are given to a selection. Derivations and proofs are integrated with applications in the text, but it is not necessary to read every line of algebra in order to find out when a method is appropriate or how to implement it. The book is readable and clearly written and would be a valuable addition to every statistician's library.

Contents:
1. Introduction
2. Estimating functions
3. Numerical algorithms
4. Working with roots
5. Methodologies for root selection
6. Artificial likelihoods and estimating functions
7. Root selection and dynamical systems
8. Bayesian estimating functions

Statistical problems often require the solution of estimation equations to obtain estimates. Sometimes multiple solutions occur, the equations exhibit nonlinearity and iterative solution methods are called for. This text provides a nicely written survey of this specialist field, written from mathematical and computing points of view, and makes use of MAPLE, MATLAB and Mathematica programming methods. It is an excellent book for its intended audience. Minor point: On page iv we see "The moral rights of the author have been asserted"; this statement is not further explained. More information can be found at
www.hmso.gov.uk/acts/acts1988/Ukpga_19880048_en_5.htm#mdiv77 and www.ariadne.ac.uk/issue4/copyright/ .

Contents:
Introduction: Why Bayesian nonparametrics – An Overview and Summary
1. Preliminaries and the finite dimensional case
2. M(X) and priors on M(X)
3. Dirichlet and Polya Tree process
4. Consistency theorems
5. Density estimation
6. Inference for location parameter
7. Regression problems
8. Uniform distribution on infinite-dimensional spaces
9. Survival analysis - Dirichlet priors
10. Neutral to the right priors
11. Exercises

Readership: Academic (postgraduate students and researchers in Bayesian nonparametrics)

The style of the book is well summarized in the following quotations: "This monograph provides a systematic, theoretical development of the subject." and "We view this book as an introduction to the theoretical aspects of the topic at a graduate level. There is no coverage of the important aspect of computations." The treatment is quite formally mathematical, and not for the faint-hearted; the material largely comprises definitions, propositions, theorems and proofs. A strong background in mathematics, including the measure-theoretic approach to probability is required.
The Introduction, "Overview and Summary", is unusually long and detailed, and does its job very well. In the words of a current television advertisement for a weather-proofing liquid, "It does exactly what it says on the tin". The prerequisites are briefly outlined in Chapter 1: these include metric spaces (compactness, separability, Borel sigma-algebras and weak convergence), posterior consistency and robustness, and non-regular theory. Chapter 2 is mainly concerned with existence theorems for priors on an infinite-dimensional parameter space, typically the set of all probability measures on the sample space. Chapter 3 gives detailed coverage of the Dirichlet process prior, essential study for this area. The Polya Tree process is also described; this is a more recent framework that yields a richer class of priors at some cost in mathematical tractability. Chapter 4 contains a bunch of theorems relating to posterior consistency, and the results are applied in subsequent chapters. Chapter 5 ("this rather technical chapter") focuses on priors for density estimation and their consistency. Chapter 6 gives results concerning posterior consistency for the location parameter in a semi-parametric setting and proposes suitable Polya Tree priors. The related set-up of a linear regression model is tackled in Chapter 7, in which the coefficients and error distribution are to be estimated. The question in Chapter 8 is how to construct a non-informative prior on an infinite-dimensional space; a starting point is the Jeffreys' prior for a finite-dimensional parameter. Survival analysis, with right-censored observed failure times, is addressed in Chapters 9 and 10, where appropriate priors are suggested and investigated. Some miscellaneous exercises are presented in Chapter 11 and the index contains references to subjects and authors in roughly equal amounts.
The book will find a place as essential study for researchers in this modern area of statistics. It is well written, the signposts are clearly displayed throughout, and the literature appears to be well documented.

Contents:
PART l: Fundamentals of Bayesian Inference
1. Background
2. Single-parameter models
3. Introduction to multipararameter models
4. Large-sample inference and frequency properties of Bayesian inference
PART II: Fundamentals of Bayesian Data Analysis
5. Hierarchical models
6. Model checking and improvement
7. Modeling accounting for data collection
8. Connections and challenges
9. General advice
PART III: Advanced Computation
10. Overview of computation
11. Posterior simulation
12. Approximations based on posterior modes
13. Special topics in computation
PART IV: Regression Models
14. Introduction to regression models
15. Hierarchical linear models
16. Generalized linear models
17. Models for robust inference
18. Mixture models
19. Multivariate models
20. Nonlinear models
21. Models for missing data
22. Decision analysis

Readership: Graduate and advanced undergraduate students of statistics, and practitioners and researchers in applied statistics

This second edition [First edition 1995; Short Book Reviews, Vol. 16, p. 5] expands on the already comprehensive first edition with the addition of material on model checking, data collection and computation; new chapters on nonlinear models and decision analysis; an appendix illustrating computation using the software packages R (general statistics) and Bugs (Bayesian inference); and a remarkable collection of applied examples from the authors' recent research. The book concentrates on appli-cations and computing on the view that the field has moved beyond philosophical debate about foundations of inference.
Some statistical maturity is expected of the audience, and in return the reader is rewarded with exceptionally cogent and precise discussions of Bayesian inference, data analysis and computing. It is a pleasure to Iearn from this book. It is a masterful consolidation of recent developments, especially as a snapshot of the many rapidly expanding areas of application. This is an essential reference for current and future Bayesian data analysts.

Reviewer:
Institute	Katholieke Universiteit Leuven
Place	Leuven, Belgium
Name	J.L. Teugels

Reviewer:
Institute	Imperial College of Science,Technology and Medicine
Place	London, U.K.
Name	M.J. Crowder

Reviewer:
Institute	Imperial College of Science,Technology and Medicine
Place	London, U.K.
Name	M.J. Crowder

Reviewer:
Institute	Imperial College of Science,Technology and Medicine
Place	London, U.K.
Name	D.J. Hand

Reviewer:
Institute	Imperial College of Science,Technology and Medicine
Place	London, U.K.
Name	D.J. Hand

Title	THE ANALYSIS OF TIME SERIES. AN INTRODUCTION, 6th edition.
Author	C. Chatfield.
Publisher	Boca Raton, Florida: Chapman and Hall/CRC Press, 2003, pp. xiii + 333.

Reviewer:
Institute	ETH-Zürich
Place	Zürich, Switzerland
Name	P.A.L. Embrechts

Reviewer:
Institute	Technische Universität Berlin
Place	Berlin, Germany
Name	F. Esche

Reviewer:
Institute	Limburgs Universitair Centrum
Place	Diepenbeek, Belgium
Name	N.D.C. Veraverbeke

Reviewer:
Institute	University of Western Ontario
Place	London, Canada
Name	D.R. Bellhouse

Reviewer:
Institute	University of Toronto
Place	Toronto, Canada
Name	D.F. Andrews

Reviewer:
Institute	Nuffield College
Place	Oxford, U.K.
Name	D.R. Cox

Reviewer:
Institute	Imperial College of Science, Technology and Medicine
Place	London, U.K.
Name	F.H. Berkshire

Reviewer:
Institute	Queen's University
Place	Kingston, Canada
Name	M.R. Murty

Reviewer:
Institute	Macquarie University
Place	Sydney, Australia
Name	J.R. Leslie

Reviewer:
Institute	Brookfield
Place	U.S.A.
Name	C.A. Fung

Reviewer:
Institute	Johns Hopkins University
Place	Baltimore, U.S.A.
Name	T.A. Louis

Reviewer:
Institute	University College London
Place	London, U.K.
Name	V.S. Isham

Reviewer:
Institute	University of Wisconsin
Place	Madison, U.S.A.
Name	N.R. Draper

Reviewer:
Institute	University of Waterloo
Place	Waterloo, Canada
Name	R.W. Oldford

Reviewer:
Institute	GlaxoSmithKline
Place	Collegeville, U.S.A.
Name	V.V. Fedorov

Reviewer:
Institute	Sofia University
Place	Sofia, Bulgaria
Name	B.I. Penkov

Reviewer:
Institute	University of Exeter
Place	Exeter, U.K.
Name	W.J. Krzanowski

Reviewer:
Institute	University of St. Andrews
Place	St. Andrews, U.K.
Name	C.D. Kemp

Reviewer:
Institute	University of Berne
Place	Berne, Switzerland
Name	J. Hüsler

Reviewer:
Institute	University of Aarhus
Place	Aarhus, Denmark
Name	S. Asmussen

Reviewer:
Institute	University of Newcastle
Place	Newcastle, Australia
Name	G.C. Goodwin

Reviewer:
Institute	London South Bank University
Place	London, U.K.
Name	S. Starkings

Reviewer:
Institute	London School of Economics
Place	London, U.K.
Name	S. Powell

Reviewer:
Institute	Technology and Medicine
Place	London, U.K.
Name	D.J. Balding

Reviewer:
Institute	University of Southampton
Place	Southampton, U.K.
Name	P. Prescott

Reviewer:
Institute	Pennsylvania State University
Place	University Park, U.S.A.
Name	T.P. Hettmansperger

Reviewer:
Institute	University of Southampton,
Place	Southampton, U.K.
Name	P. Prescott

Reviewer:
Institute	University of Kent
Place	Canterbury, U.K.
Name	B.J.T. Morgan

Reviewer:
Institute	University of Kent,
Place	Canterbury, U.K.
Name	B.J.T. Morgan

Reviewer:
Institute	MRC Biostatistics Unit
Place	Cambridge, U.K.
Name	V.T. Farewell