Short Book Reviews On-Line 1994

200% OF NOTHING. An Eye-Opening Tour Through the Twists and Turns of Math Abuse and Innumeracy.	A.K. Dewdney.
MEDICAL STATISTICS. A Commonsense Approach, 2nd edition.	M.J. Campbell and D. Machin.
DRUG SAFETY ASSESSMENT IN CLINICAL TRIALS.	G. Sogliero-Gilbert (Ed.).
STATISTICS FOR THE 21st CENTURY.	J.W. Duncan and A.C. Gross.
GEOMETRICAL AND STATISTICAL METHODS OF ANALYSIS OF STAR CONFIGURATIONS: DATING PTOLEMY'S ALMAGEST.	A.T. Fomenko, V.V. Kalashnikov and G.V. Nosovsky.
HUMAN RELIABILITY AND SAFETY ANALYSIS DATA HANDBOOK.	D.I. Gertman and H.S. Blackman.
STATISTICAL MODELS FOR CAUSAL ANALYSIS.	R.D. Retherford and M.K. Choe.
PROBLEMS AND SNAPSHOTS FROM THE WORLD OF PROBABILITY.	G. Blom, L. Holst and D. Sandell.
HANDBOOK OF SMALL DATA SETS. D.J. Hand, F. Daly,	A.D. Lunn, K.J. McConway and E. Ostrowski (Eds.).
OUTLIERS IN STATISTICAL DATA, 3rd Edition.	V. Barnett and T. Lewis.
INTRODUCTORY STATISTICS FOR BIOLOGY STUDENTS.	T.A. Watt.
STATISTICAL THEORY, 4th edition.	B.W. Lindgren.
ELEMENTARY PROBABILITY.	D. Stirzaker.
PROBABILITY.	J. Pitman.
PROBABILITY.	A.F. Karr.
STATISTICAL MODELS IN EPIDEMIOLOGY.	D. Clayton and M. Hills.
MODELLING SURVIVAL DATA IN MEDICAL RESEARCH.	D. Collett.
MODELS FOR REPEATED MEASUREMENTS.	J.K. Lindsey.
RANDOM COEFFICIENT MODELS.	N.T. Longford.
NUMBER-THEORETIC METHODS IN STATISTICS.	K.T. Fang and Y. Wang
MODELLING COVARIANCES IN LATENT VARIABLES USING EQS.	G. Dunn, B. Everitt and A. Pickles.
PREDICTIVE INFERENCE: AN INTRODUCTION.	S. Geisser.
COMPUTER INTENSIVE STATISTICAL METHODS. Validation, Model Selection and Bootstrap.	J.S. Urban Hjorth.
MEASUREMENT, REGRESSION AND CALIBRATION.	P.J. Brown.
MODEL-FREE CURVE ESTIMATION.	M.E. Tarter and M.D. Lock.
INFERENCE AND ASYMPTOTICS.	O.E. Barndorff-Nielsen and D.R. Cox.
EFFICIENT AND ADAPTIVE ESTIMATION FOR SEMIPARAMETRIC MODELS.	P.J. Bickel, C.A.J. Klaassen, Y. Ritov and J.A. Wellner.
CASE STUDIES IN TIME SERIES ANALYSIS.	Z. Xie.
INTRODUCTION TO MULTIPLE TIME SERIES ANALYSIS, 2nd edition.	H. Lütkepohl.
ELEMENTS OF MULTIVARIATE TIME SERIES ANALYSIS.	G.C. Reinsel.
PROBABILITY THEORY, AN ANALYTIC VIEW.	D.W. Strook.
STATISTICS FOR THE ENVIRONMENT.	V. Barnett and K.F. Turkman (Eds.).
CASE STUDIES IN BAYESIAN STATISTICS.	C. Gatsonis, J.S. Hodges, R.E. Kass and N.D. Singpurwalla (Eds.).
BIOMETRIE: MODELISATION DE PHENOMENES BIOLOGIQUES.	R. Tomassone, C. Dervin and J.-P. Masson.
MARKOV CHAINS AND STOCHASTIC STABILITY.	S.P. Meyn and R.L. Tweedie.
OPERATOR-LIMIT DISTRIBUTIONS IN PROBABILITY THEORY.	Z.J. Jurek and J.D. Mason.
HANDBOOK FOR DIGITAL SIGNAL PROCESSING.	S.K. Mitra and J.F. Kaiser (Eds.).
THE PRACTICE OF OPERATIONAL RESEARCH.	G. Mitchell.
MODEL SOLVING IN MATHEMATICAL PROGRAMMING.	H.P. Williams.
FUZZY SETS AND INTERACTIVE MULTIOBJECTIVE OPTIMIZATION.	M. Sakawa.
QUEUING THEORY IN MANUFACTURING SYSTEMS ANALYSIS AND DESIGN.	H.T. Papadopoulos, C. Heavey and J. Browne.
TOTAL QUALITY MANAGEMENT PROCESSES: A Systematic Approach.	G.K. Kanji and M. Asher.
QUALITY CONTROL IN ANALYTICAL CHEMISTRY, 2nd edition.	G. Kateman and L. Buydens.
THE CRAFT OF DECISION MODELLING.	P. Rivett.
BAYESIAN THEORY.	J.M. Bernardo and A.F.M. Smith.
THE ESSENTIALS OF PROBABILITY.	R. Durrett. Belmont,
THE INVERSE GAUSSIAN DISTRIBUTION. A Case Study in Exponential Families.	V. Seshadri.
KENDALL'S ADVANCED THEORY OF STATISTICS. VOLUME I: DISTRIBUTION THEORY, 6th edition.	A. Stuart and J.K. Ord.
A HANDBOOK OF STATISTICAL ANALYSES USING S-PLUS.	B.S. Everitt.
STATISTICAL METHODS FOR SPC AND TQM.	D. Bissell.
STATISTICAL QUALITY CONTROL WITH MICROCOMPUTER APPLICATIONS.	L.E. Shirland.
GENERAL PATTERN THEORY: A MATHEMATICAL STUDY OF REGULAR STRUCTURES.	U. Grenander.
DESIGN AND ANALYSIS OF EXPERIMENTS. VOLUME 1. Introduction to Experimental Design.	K. Hinkelmann and O. Kempthorne.
AGRICULTURAL FIELD EXPERIMENTS: Design and Analysis.	R.G. Petersen.
EXPERIMENTAL RESEARCH DESIGN AND ANALYSIS. A Practical Approach for the Agricultural and the Natural Sciences.	A.R. Hoshmond.
ANALYSIS OF LONGITUDINAL DATA.	P.J. Diggle, K.-Y. Liang, and S.L. Zeger.
ADVANCED LINEAR MODELS, Theory and Applications.	S.-G. Wang and S.-C. Chow.
LOGISTIC REGRESSION WITH MISSING VALUES IN THE COVARIATES.	W. Vach.
NONPARAMETRIC REGRESSION AND GENERALIZED LINEAR MODELS. A Roughness Penalty Approach.	P.J. Green and B.W. Silverman
PERMUTATION TESTS. A Practical Guide to Resampling Methods for Testing Hypotheses.	P. Good.
DISTANCE SAMPLING: ESTIMATING ABUNDANCE OF BIOLOGICAL POPULATIONS.	S.T. Buckland, D.R. Anderson, K.P. Burnham and J.L. Laake.
NUMERICAL SOLUTION OF SDE THROUGH COMPUTER EXPERIMENTS.	P.E. Kloeden, E. Platen and H. Schurz.
MACHINE LEARNING, NEURAL AND STATISTICAL CLASSIFICATION.	D. Michie, D.J. Spiegelhalter and C.C. Taylor (Eds.).
STATISTICAL APPLICATIONS USING FUZZY SETS.	K.G. Manton, M.A. Woodbury and H.D. Tolley.
MEASURE THEORY.	J.L. Doob.
HILBERT SPACE METHODS IN PROBABILITY AND STATISTICAL INFERENCE.	C.G. Small and D.L. McLeish.
PROBABILITY WITH A VIEW TOWARD STATISTICS. Volumes I and II.	J. Hoffmann-Jörgensen.
AN INTRODUCTION TO STOCHASTIC MODELING. Revised edition.	H.M. Taylor and S. Karlin.
NUMERICAL METHODS FOR STOCHASTIC PROCESSES.	N. Bouleau and D. Lepingle.
SIMULATION AND CHAOTIC BEHAVIOUR OF á-STABLE STOCHASTIC PROCESSES.	A. Janicki and A. Weron.
TIME SERIES ANALYSIS FORECASTING AND CONTROL, 3rd edition.	G.E.P. Box, G.M. Jenkins and G.C. Reinsel.
TIME-SERIES MODELLING OF WATER RESOURCES AND ENVIRONMENTAL SYSTEMS.	K.W. Hipel and A.I. McLeod,
MULTIVARIATE SYSTEMS AND FILTER BANKS.	P.P. Vaidyanathan.
TOPICS IN CONTROL THEORY.	H.W. Knobloch, A. Isidori and D. Flockerzi.
GAME THEORY AND STRATEGY.	P.D. Straffin.

Contents:
1. Innumeracy and math abuse
2. Statistics and damned lies
3. The mathematics of advertising
4. Intelligent dice
5. The law of zero return
6. Caveat emptor
7. The government figures
8. Living with risk
9. Gee-whiz media math
10. The tip of the iceberg
11. Everybody is a mathematician
12. Street math

This is a non-technical collection of mathematical and statistical horror stories. The author wrote a popular "Mathematical Recreations" column in Scientific American for eight years. Many of the examples in this book were sent in by his readers, called "abuse detectives", members of the public who spotted abuses of mathematics and statistics in the mainly U.S. media, in advertising, in politics or whatever. Some examples will be familiar to statisticians in general terms from earlier books such as D. Huff's How to Lie with Statistics.

Contents:
1. Uses and abuses of medical statistics
2. Design
3. Probability and decision making
4. Data description
5. From sample to population
6. Statistical inference
7. Correlation and linear regression
8. The randomized controlled trial
9. Designed observational studies
10. Common pitfalls in medical statistics

Readership: Consumers of medical research, biomedical investigators, some medical practitioners

The authors' goal is to present statistical principles for the critical reader of the medical literature by emphasizing design over analysis and interpretation over technique. The presentation is highly structured. Every chapter starts with a summary and ends with a checklist called "points when reading the literature". The book is rich in examples from the literature. Each and every one is labelled, "Example from the Literature". One suspects that all the sign-posting in this book is to remind the reader how relevant statistical science is to medical research. The feeling is heightened by frequent admonitions to "see your statistician early". All but the simplest formulae and calculations are segregated in an appendix which gives worked examples of eighteen procedures from, "Notation and Hints on Calculation", to "The Kaplan-Meier Curve and Logrank Test". Whatever techniques the authors included, a statistical reviewer could fuss. For me, the section on non-independent residuals was murky and the omission of the Cox regression regrettable. As a book that explicates statistics for the critical medical consumer who does not want/need to worry about technique, this one seems to meet its authors' objectives. However, I do not believe that statistical principles always can conform to common sense.

Contents:
1. Preclinical drug safety evaluation, J.T. Mayne
2. Adverse drug events in clinical trials, D.S. Kirby
3. Laboratory testing in clinical trials, D.S. Kirby
4. Drug research in the elderly, P.M. Hooymans and R. Janknegt
5. Drug assessment in critical illness, M.I. Bowden and J.F. Bion
6. Laboratory data in multicenter trials: Monitoring, adjustment, and summarization, L.K. Oliver and C. Chuang-Stein
7. The Genie Score: A multivariate assessment of laboratory abnormalities, G. Sogliero-Gilbert, L. Zubkoff-Schulz and N. Ting
8. Laboratory parameters and drug safety, N.E. Pitts
9. A unified approach to the analysis of safety data in clinical trials, C. Chuang-Stein and N.R. Mohberg
10. The use of hazard functions in safety analysis, D. Salzburg
11. Meta-analysis of drug safety data, G.G. Koch, J.E. Schmid, J.M. Begun and W.C. Maier
12. Design and analysis considerations for safety data, particularly adverse events, K.E. Peace
13. Clinical trial adverse drug experience reporting requirements in the major countries: One manufacturer's approach, M.W. Talbott and E.D. Kelso
14. Safety surveillance, N.E. Pitts
15. Post-marketing surveillance: Applications and limitations, with special reference to the fluoroquinolines, R. Janknegt and Y.A. Hekster

Readership: Statisticians who are or will be involved in clinical trials of new drugs or in post-marketing surveillance

This edited book may be the first text devoted entirely to the subject of drug safety assessment. The contributing authors have broad and international experience in the assessment of safety, and hence are well-suited to comment on this specialized area. The text covers both standard and well-established approaches to drug safety assessment and also on some issues, for example the combining of the evidence across multiple studies, which modern computing and statistical methods have repopularized. This represents a useful text for statisticians about to embark upon a career in the pharmaceutical industry, as well as for others wishing to learn the key issues in the safety assessment of drugs in clinical trials.

Contents:
1. Nature of problem
2. "SWOT" analysis for statistics
3. Elements of the statistical system – The stakeholders and interactions
4. A cast study: Health care statistics
5. Fundamentals - national income accounting
6. Global interaction
7. Recommendations. National accounts and the SNA
8. Recommendations. International statistics
9. Concluding remarks
APPENDIX: Guidelines for bibliography development
The general accounting office's role in federal statistical accounting and related reports on statistical policy issues
Errors in GDP estimates
Statistical budgets
Services and invisible methodology problems

Readership: Those involved, or at least interested, in statistics for decision-making at national supra-national level, or at the level of commercial and other organizations

It is clear that we have been through one information revolution with the development of computers. It is also clear that we are just at the beginning of a much greater one, in which the direct linkage of computers will make things possible of which we scarcely dream though we are also in some danger of being smothered by information. Our classification methods have not been updated at anything like the same speed as the technology; thus official statistics are much better at analyzing heavy industry than services, even though the latter is beginning to out-weigh the former. In this book such questions are discussed. The details are almost entirely based on the situation in the United States.

Contents:
1. Some concepts of astronomy and history of astronomy
2. Star catalogue of the Almagest. Preliminary analysis
3. The attempts to date the Almagest with the help of the simplest procedures, and why they fail
4. Who is who?
5. Analysis of systematic errors in stellar configurations
6. Statistical properties and the accuracy of the catalogue of the Almagest
7. Dating the star catalogue of the Almagest
8. Inclination of the ecliptic in the Almagest
9. Dating other medieval catalogues
10. A date for the Almagest for coverings of stars and lunar eclipses.
Addendum: Problems and hypotheses connected with dating the Almagest

This is a historical detective story about the existence and veracity of Ptolemy; either he was ancient and a cheat, or medieval and an imposter. The authors search for patterns to explain anomalies in a purportedly 1800 year old star catalogue, by a lot of data mining and data rejection followed by least- squares fitting; it seemed to me a pity that robust methods were not used to derive less subjective and more powerful conclusions. The authors settle for the imposter hypothesis, but it only emerges in the addendum that this was developed 60 years ago by Morozov.
The reader will have to do some detective work too; several crucial numerical quantities are mis-printed, and the tables have no titles nor explanatory captions although the figures have very clear captions. There should be some good student projects in this material.

Contents:
1. Introduction and background to human reliability analysis
2. Conducting human reliability analysis
3. Formal methods for estimating human reliability
4. HRA fault and event trees
5. Existing data sources and data bank
6. HRA: A case study for nuclear processing facility design
7. HRA case study for a nuclear power plant; containment venting procedure
8. Relation of HRA to system safety and system performance
9. Simulators and simulation as a tool for evaluating human reliability
10. Organisational factors and human reliability
11. Outstanding issues
12. Behavioral mechanisms underlying human error
13. The problem: Representation of errors of commission(cognitive error) in PRA
14. HRA and the impact of emerging hardware and software technologies

Readership: Engineers, risk analysts, reliability analysts, particularly in the nuclear, process control and transportation industries

The book presents a summary of current methods, techniques, data and concepts in Human Reliability Assessment (HRA). It argues for a systematic and quantifiable approach to human error. This will enable a range of improvements to complex man-machine systems which in turn will reduce risk. The authors specifically recognize both cognitive and organizational factors as contributing to risk and note that, unlike mechanical failure which is all-at-once, human perform-ance may degrade slowly over time. This is a highly specialized book, really suited only to its intended audience. For which, it provides a far-ranging overview of the topic and a rich source of references.

Contents:
1. Bivariate linear regression
2. Multiple regression
3. Multiple classification analysis
4. Path analysis
5. Logit regression
6. Multinomial logit regression
7. Survival models, Part 1: Life tables
8. Survival models, Part 2: Proportional hazard models
9. Survival models, Part 3: Hazard models with time dependence

Readership: Social science and biomedical professionals, graduate or advanced undergraduate students

This book presents a useful compendium of causal analysis techniques. Written in a clear, crisp style, it covers the field briskly, assuming a basic knowledge of statistics and trimming some of the back-ground which would require complicated mathematical argument, for example the derivation of sampling distributions and standard errors. By being very focused and precise, the authors manage to convey a great deal about causal models and the process of model building. Practitioners will find the exposition of underlying assumptions, and the effects of violating them, particularly helpful.
Although it can be used on its own, this book comes into its own as a supplementary or reference text. It is well illustrated with examples drawn from an actual study, giving it an unusually consistent feel. There are sample computer programs too, for those with access to BMDP, LIMDEP and SAS.

Contents:
1. Welcoming problems
2. Basic probability theory I
3. Basic probability theory II
4. Topics from early days I
5. Topics from early days II
6. Random permutations
7. Miscellaneous I
8. Poisson approximation
9. Miscellaneous II
10. Random walks
11. Urn models
12. Cover times
13. Markov chains
14. Patterns
15. Embedding procedures
16. Special topics
17. Farewell problems

The authors state that "from the collection of 125 examples, the 'problems' are selected on the basis of their elegance and utility whereas the 'snap-shots' are intended to provide a quick overview of topics in probability." One of the pleasures in teaching a course on probability lies in discussing examples coming from diverse aspects of life. Such examples can be motivated by their intrinsic scientific interest, historical relevance or sheer beauty. All of these motivations are omnipresent throughout the set of examples. It is obvious that the authors have had fun in writing this book; they want to share their enthusiasm for the subject with a broad readership. An important bonus is that all problems come from discrete probability so that only minimal prerequisites are needed. In the good old tradition of Feller, Volume 1, it is indeed remarkable how many interesting excursions one can make in this discrete probabilistic world. Most of the chapters contain intellectual transitions from elementary calculations to taxing combinatorial arguments. The authors are to be congratulated on the final product. I am definitely looking forward to spending time browsing through the book when occupying an easy-chair.

Contents:
1. Introduction
2. How to use the disk
3. The data sets
4. Data structure index
Readership: Teachers of data analysis and statistics, statisticians

There are five hundred and ten real, small sets of data listed in the text, "real" meaning generated by observation or measurement, not invented; "small" meaning 0.81 per page, sorted conveniently in ASCII code on the accompanying DOS disk, described briefly, a dozen or so lines giving study context, source, sometimes the question(s) raised, and occasion-ally the original source, a problematic analysis, or a reference to a published analysis, and indexed by the title giving data structure, and by subject-matter con-text. The data structure code is initially cryptic but, one suspects, ultimately quite useful; it gives the number of independent units measured, the number of variables, roughly speaking, a classification of type of data and the filename on the disk. A scan of the titles in the data structure index or the keywords in the subject index is a visit to an exotic bazaar laden with fascinating trinkets from the storehouse of human curiosity and ingenuity. I cannot wait to use this book in my teaching.

Contents:
PART I : Basic Principles
1. Introduction
2. Why do outlying observations arise and what should one do about them?
3. The accommodation approach: Robust estimation and testing
4. Testing for discordancy: Principles and criteria
PART II : Univariate Data
5. Accommodation procedures for univariate samples
6. Specific discordancy test for outliers in univariate samples
PART III: Multivariate and Structured Data
7. Outliers in multivariate data
8. The outlier problem for structured data: Regression, the linear model and designed experiments
PART IV : Special Topics
9. Bayesian approaches to outliers
10. Outliers in time series: An important area of outlier study
11. Outliers in directional data
12. Some little-explored areas: Contingency tables and sample surveys
13. Important strands: Computer software, data studies, standards and regulations
14. Perspective

This is the third edition of this now well-known book. The review of the first edition has been sent as an example accompanying any book sent out for review by this publication; the review of the second edition appeared in Short Book Reviews, Vol. 5, p.3. Unlike many new editions, this one is a complete revision. Some items have been removed, for example outlying sub-samples, but new ones have been included or others changed substantially such as new tests for univariate data, non-linear regression, time series and directional data. To answer the question "what are outliers and what is the outlier problem", the authors quote B. Peirce (1852), "In almost every true series of observations, some are found, which differ so much from the others as to indicate some abnormal source of error not contemplated in the theoretical discussions, and the introduction of which into the investigations can only serve ... to perplex and mislead the inquirer."
Over 1000 new references have been included expanding the reference section from 19 pages in 1978 to 27 in 1984 to 45 in 1994. The authors are to be commended in locating all of these.
Each edition has ended on a questioning note, this one being "Surely the professional statistician needs to have a thorough modern understanding of how to handle outliers in statistical data."
This edition needs to be on the book shelf as a reference for anyone who analyzes data.

Contents:
1. How long is a worm?
2. Confidence intervals and computers
3. Sampling
4. Planning an experiment
5. Accounting for background variation and constructing a model
6. Analysing your results: Is there anything else?
7. Consider the treatments of the field
8. Relating one thing to another
9. What to do when the data are skewed or are ranks or scores or counts in categories
10. Summarizing data from an observational study
11. Your project
12. Preparing a report. What's it all for?

The author aims to calm the fears of biology students about statistics and to inspire in them a sense of the importance of statistical ideas. Unfortunately, like many similar books, this one is unsatisfactory. There are the all-too-common technical errors: a plot of the normal density appears to go to zero at the origin, for example, and the interpretation of confidence intervals is at best unclear and sometimes in-correct. The choice of topics is unusual: there is a strong emphasis on experimental design, reflecting the author's background in agriculture, but, surprisingly, no mention of either binomial or Poisson distributions and very little exploratory data analysis. Use of the F distribution in analysis of variance is advocated with no discussion of the required assumptions, al-though the assumptions underlying normal-theory inference in linear regression are discussed. Ultimately, I believe, the author aims for too much: the difficult concepts of classical frequentist inference cannot readily be taught at this level. An introductory text should more usefully occupy itself giving a firmer grasp of important elementary ideas.

Contents:
1. Preliminaries
2. Probability
3. Random variables
4. Expectations
5. Limit theorems
6. Some parametric families
7. Sampling and reduction of data
8. Estimation
9. Testing hypotheses
10. Analysis of categorical data
11. Sequential analysis
12. Multivariate distributions
13. Nonparametric tests
14. Linear models and analysis of variance
15. Decision theory

In the early 1960's there appeared several books on statistical theory, targeted at the more serious end of the undergraduate market, which have become standard texts for the subject. They have maintained their dominance via successive editions which ensure that emphasis, layout, style and quantity of examples reflects contemporary fashion. This new edition of Lindgren's book is a case in point. The first edition appeared when measure theory was in its ascendancy in departments of statistics and the influence of this theory has been apparent, albeit in a rather dilute form, in all successive editions. How-ever, the present edition appears to have been forged on the anvil of experience. Because of the weak mathematical background of many who major in statistics, any discussion of the foundations of probability and set theory proves indigestible and in any case is not essential to a reasonable grasp of undergraduate statistical theory. Thus the section dealing with the foundations has been purged of such material and made more accessible to current students. Another advance has been that more examples have been included; an increase of over forty or fifty percent depending on whether you believe the preface or the jacket. This is welcome as earlier editions had been a little light on examples.
A feature of such texts in the 1960's was that important or useful results were often buried in a body of text. This contrasts with the current trend of box-ing or high-lighting by shading and indenting every-thing in sight. The new edition has made some gesture in this direction by elevating useful results to the status of theorems. The overall content, whilst being extensively redrafted, has not changed in substance very much although there has been some re-ordering of the topics. Solutions to most examples are provided. The net effect of all this is to make the book more user-friendly. This edition should allow the text to continue to have currency in the undergraduate market and to maintain its status as one of the standards.

Contents:
1. Probability
2. Conditional probability and independence
3. Counting
4. Random variables: Distribution and expectation
5. Random vectors: Independence and dependence
6. Generating functions and their applications
7. Continuous random variables
8. Jointly continuous random variables
9. Markov chains

Excellent! A vast number of well-chosen worked examples and exercises guide the reader through the basic theory of probability at the elementary level. The author must have worked very hard to bring together so many good and useful problems and examples. This undoubtedly makes the book stand out from the multitude of texts now available on the subject. A novelty at this level is definitely the chapter on Markov chains, sixty pages. In order to keep to the elementary mathematical level, the author introduces the rather unusual notion of regularity, which ensures that a certain power of the transition matrix has no zero en-tries. This makes possible the proof of the theorem at this elementary level. All in all, this is an excellent text which I am sure will give a lot of pleasure to students and teachers alike.

Contents:
1. Introduction
2. Repeated trials and sampling
3. Random variables
4. Continuous distributions
5. Continuous joint distributions
6. Dependence

This elementary introduction to probability emphasizes the development of intuition through in-formal description together with many visual illustrations and worked examples. This approach is mathematically unsophisticated, but does not avoid important ideas: the normal approximation to the binomial distribution is carefully explained in Chapter 2, with emphasis on the accuracy of the approximation, not merely a valid limit statement. The interface of theory with applications is well addressed: true odds are distinguished from betting odds and frequentist and subjective interpretations are discussed and their implications integrated into the text. The leisurely pace implies that, even with well over five hundred pages, more advanced topics are not covered, among them being transformations of joint densities, generating functions and elementary stochastic processes. This book is not just a reference resource, but is designed for students to learn. It is thus regrettable that, currently available only in hardback, its major drawback is its excessive price.

Contents:
Prelude: Random Walks
1. Probability
2. Random variables
3. Independence
4. Expectation
5. Convergence of sequences of random variables
6. Characteristic functions
7. Classical limit theorems
8. Prediction and conditional expectation
9. Martingales

This book attempts to answer the question: "How much of the mathematical foundations of probability should a graduate student of statistics know?". To quote from the preface: "The major issue in writing a book on probability at this level is what to do about measure theory". The approach here is to use some concepts and results from measure theory, but avoid a general development. The selection of material is traditional; Chapters 1 to 7 work towards a climax at the three great classical limit theorems, while Chapters 8 and 9 introduce conditional expectation, in terms of minimum mean-squared error of prediction and martin-gales. The development is reasonably formal and rigorous, with many proofs worked in full, although results are usually not given in their most general form. Applications to, for example, stochastic integration and maximum-likelihood estimation, are briefly introduced, but Karr regrets that space restrictions did not
allow a full discussion of applications.

Contents:
PART I : Probability Models and Likelihood
1. Probability models
2. Conditional probability models
3. Likelihood
4. Consecutive follow-up intervals
5. Rates
6. Time
7. Competing risks and selection
8. The Gaussian probability model
9. Approximate likelihoods
10. Likelihood, probability and confidence
11. Null hypothesis and p-values
12. Small studies
13. Likelihoods for the rate ratio
14. Confounding and standardization
15. Comparison of rates within strata
16. Case-control studies
17. Likelihoods for the odds ratio
18. Comparison of odds within strata
19. Individually matched case-control studies
20. Tests for trend
21. The size of investigations
PART II: Regression Models
22. Introduction to regression models
23. Poisson and logistic regression
24. Testing hypotheses
25. Models for dose-response
26. More about interaction
27. Choice and interpretation of models
28. Additivity and synergism
29. Conditional logistic regression
30. Cox's regression analysis
31. Time-varying explanatory variables
32. Three examples
33. Nested case-control studies
34. Gaussian regression models
35. Postscript

Statistics as practiced by epidemiologists is somewhat specialized. Often the epidemiologists learn statistics through the classical development of the normal theory model with extensions of the methods to epidemiologic problems. This book reverses this order and hence will appeal to specialists in epidemiology. Understanding of the material will be enhanced by prior exposure to statistical methods, but extensive mathematical preparation is not required. The book will work best as a text for an instructed course or for self- teaching. It is readable and contains many, solved, exercises for practice.

Contents:
1. Survival analysis
2. Some non-parametric procedures
3. Modelling survival data
4. The Weibull model for survival data
5. Model checking in the proportional hazards model
6. Some other parametric models for survival data
7. Time-dependent variables
8. Interval-censored survival data
9. Sample size requirements for a survival study
10. Some additional topics
11. Computer software for survival data
12. Time-dependent variables
13. Interval-censored survival data
14. Sample size requirements for a survival study
15. Some additional topics
16. Computer software for survival analysis

Readership: Statisticians, numerate scientists and clinicians, undergraduate students

Anyone wanting an easy introduction to per-forming a survival analysis should buy this book. The author gives an outstanding lucid account of both the parametric and non-parametric approaches to survival analysis.
The first five chapters are devoted to Cox's proportional hazards model, and models based on the Weibull distribution. All aspects of modelling, from model definition and selection to residual analysis, are discussed. Sufficient mathematical detail is given for the motivation behind the techniques to be grasped and references are given for further reading. As each concept is introduced, the author gives an example that illustrates it precisely. Thirteen sets of data are analyzed, some of them by several methods, which allow the methods to be compared and contrasted. A particularly attractive feature is the detailed interpretation of the results of each of the analyses.
In the next chapters, other survival models are discussed: accelerated failure-time models, models with time-dependent covariates, non-proportional hazards, multistate models and models for interval- censored-survival data. The chapter on sample size requirements and the references therein will be useful to those planning a survival study. The final chapter compares the various computer programs available for survival analysis.
This book complements Cox, D.R. and Oakes, D. (1984) Analysis of Survival Data, [Short Book Reviews, Vol. 4, p.35] and, of course, incorporates many of the developments that have occurred over the last ten years. Although the title suggests that it is intended for medical applications, it will be useful to practitioners in any field. It would be a suitable textbook for an undergraduate course in survival analysis. It should do much to encourage use of these powerful techniques in other areas. The book is highly recommended.

Contents:
PART I : Introduction
1. Basic concepts
2. Fundamentals of modelling
PART II : Normal Distribution Models
3. Heterogeneous populations
4. Longitudinal studies
PART III: Models for Categorical Data
5. Overdispersion
6. Longitudinal discrete data
PART IV : Models for Duration of Data
7. Frailty
8. Event histories

Readership: Research statisticians in agriculture, medicine, economics and psychology, and consulting statisticians

After a brief introductory chapter the book gives a sixty-six page outline of statistical modelling in general, before focusing on the particular problems of repeated measurements. An attractive feature is the book's wide scope; it includes discussions of categorical data as well as normal distribution models and duration data. It has many real examples, with complete listings of the data. The list of 1382 references has an associated key classifying the contents of each article according to twenty-one categories, which makes the book a valuable resource for anyone who is thinking of under-taking research in this area.
The author describes his intent as being "to provide methods and examples which may form a basis from which a research worker can proceed", and says that the book is aimed primarily at applied statisticians, but it is also suitable as a graduate text. I think that the author has hit his target. Given the wide range of topics covered, the author necessarily had to omit or skip a number of areas. These include inference, for which he relies on deviances or likelihood ratios, extensions to multivariate responses, and design of repeated measurements studies. Whether one thinks that some of these should be included at the expense of the topics he has covered is a matter of taste. My own view is that this volume is a valuable contribution to the literature on repeated measurements.

Contents:
1. Clustered observations
2. Analysis of covariance with random effects
3. Examples. Random effects models
4. Random regression coefficients
5. Examples using random regression coefficient models
6. Multiple levels of nesting
7. Factor analysis and structural equations
8. GLM with random coefficients

Readership: Statisticians and researchers who have to handle clustered observations, predominantly, but not solely, from the social sciences

I would describe this as an applied statistics book: it illustrates the ideas through many real examples but does not shy away from the necessary mathematics. The examples bring home the importance and appropriateness of random coefficient models and are highly convincing for the author's goal of rehabilitating variation as an entity of substantive interest in the social sciences. In particular, the book has three aims: to describe methods for analyzing clustered observations, to provide examples where between cluster variation is of substantive interest, and to provide a balanced presentation of the advantages and limitations of these methods. It seems to me to be successful in achieving these aims.
A basic knowledge of linear algebra and statistics, including maximum likelihood estimation and the multivariate normal distribution, is assumed. Most of the chapters conclude with a bibliographical notes sec-tion which provide good entry points to the literature. I did not spot many errors, though "accommodate"
is consistently misspelt in the preface and elsewhere.

Contents:
1. Introduction to the number-theoretic method
2. Numerical evaluation of multiple integrals in statistics
3. Optimisation and its application in statistics
4. Representative points of a multivariate distribution
5. Experimental design and design of computer experiments
6. Some applications in statistical inference

This intriguing text is a rigorous mathematical development of results using number-theoretic methods for numerical integration, optimization and statistical simulation, developed over many years by the authors. The method finds a set of points, called an NT-net, which achieves uniform scatter over a multidimensional unit cube, in the sense of having small discrepancy, a measure defined by the authors. Use of good-lattice-point sets is shown to produce much smaller errors in numerical integration than use of either equilattice points or Monte-Carlo simulation. The method is illustrated using various multivariate moment and probability integrations. A sequential algorithm involving iteratively-generated NT-nets, de-fined over changing domains, is introduced to solve optimization problems including estimation with maxi-mum likelihood and non-linear regression. The experimental design chapter considers the use of good-lattice-point sets to determine uniform designs in which the number of design points is very small com-pared with the number of factor level combinations. Models including interaction terms are fitted using stepwise regression since, for these designs, there may be fewer observations than model parameters. The chapter includes a useful up-to-date section on designs for computer experiments. The final chapter deals with robust multivariate estimation and tests for multinormality and sphericity. Exercises are given at the end of each chapter. The book offers an interesting opportunity to learn about a particular optimization method applied to a range of statistical problems.

Contents:
Overview
1. Prelude: The ideas of covariance and of covariance structure
2. Writing a simple EQS program: Learning the language
3. Statistical modelling in EQS
4. Confirmatory factor analysis models
5. Multitrait-multimethod and multiple indicator multiple cause models
6. Models for longitudinal data
7. Simultaneous analysis for two or more groups
8. Common practical problems

This book is well planned and well written. It will be essential reading for EQS users. All the models discussed share a latent variable structure illustrated by a path diagram, for example, confirmatory factor analysis. EQS is a program for fitting these models, marketed by BMDP since 1992. The program is described in this book, and throughout there are displays giving EQS codes and results. A strength of the book is its discussion of thorny problems, for example, hitting parameter boundaries, under-identified models. Presenting formulae for expressions like Var(X+Y), the authors write: "We ... are content to let the reader use them as given." This category of reader will be thrown by a notational error in Display 1.12, which sets up a path diagram for a factor analysis model. Such readers are likely to find the more complex models of this book hard to fit and understand without expert help. It is to be hoped that this book will encourage non-statisticians to seek statistical advice.

Contents:
1. Introduction
2. Non-Bayesian predictive approaches
3. Bayesian prediction
4. Selecting a statistical model and predicting
5. Problems of comparison and allocation
6. Perturbation analysis
7. Process control and optimization
8. Screening tests for detecting a characteristic
9. Interim analysis and sampling curtailment

This is an excellent introduction to statistical prediction written by one of the leading proponents of the predictive approach to statistics. The text is clearly and succinctly written and covers pre-diction from many perspectives including the Bayesian, likelihood, and cross-validatory or sample reuse approaches. It is currently the most complete and useful introductory survey of predictive methodology and provides a good starting point for anyone interested in learning or using predictive methods. I recommend it wholeheartedly.

Contents:
1. Prelude
2. Computer-intensive philosophy
3. Cross validation
4. Validation of time series problems
5. Statistical bootstrap
6. Further bootstrap results
7. Computer-intensive applications

It may seem somewhat surprising that Chapman and Hall have chosen to publish this book so soon after B. Efron and R.J. Tibshirani's comprehensive monograph An Introduction to the Bootstrap [Short Book Reviews, Vol. 14, p.4]. A substantial proportion of Hjorth's work, especially the descriptive material on the boot-strap in Chapters 5 and 6, covers material considered in greater depth by Efron and Tibshirani, and the mathematical level is not very different. But this book leans heavily towards applications. It gives both a clear introductory discussion of the computer-intensive philosophy and a number of interesting examples relating to problems in different fields, including meteorology, economics and road safety analysis. These will provide useful insights of computer-intensive methodology and the bootstrap for non-statisticians. The book is written in an engaging, entertaining style and is well worth a look, though the statistician wishing to learn in a structural way about computer-intensive methods for statistical inference will prefer the book of Efron and Tibshirani.

Contents:
1. Introduction
2. Simple linear regression
3. Multiple regression and calibration
4. Regularized multiple regression
5. Multivariate calibration
6. Regression on curves
7. Non-linearity and selection
8. Pattern recognition

The author states that "this book has been designed primarily as a research monograph for a range of regression problems...". This is an advanced text on regression, with an empahsis upon calibration. Readers will need a good knowledge of regression at say the level of G.A.F. Seber's Linear Regression Analysis to take advantage of this book. Some of the topics covered include profile likelihood, ridge regression, partial least squares, Bayesian prediction and calibration, and pattern recognition.
As calibration is a major theme of this text, I was disappointed with the way in which calibration was introduced in Chapter 2. The author seems to be in too much of a hurry to get to the more advanced mathematics. The distinction between the two cases, which he calls natural or random calibration and controlled calibration, should have been treated at much greater length. At least one real example should have been included.
The title of Chapter 7 suggests that it is on non-linear regression and variable selection in multiple regression. In fact it is on the problem of prediction from spectra, such as near infra-red spectra. This chapter is one of the easier ones to read; it is well illustrated and should be useful to workers in this area.
Subjects which are not covered, or which receive only very limited comment, include 'errors in variables' which is relevant to the problem of calibrating several instruments which are supposed to be measuring the same quantity, the use of splines, and non-linear regression.

Contents:
1. Introduction to curve estimation
2. Generalized representation
3. Series and kernel-based density estimation
4. Optimizing density estimates
5. Mixture decomposition applications
6. Curve estimation approaches to model and transformation selection
7. Threshold parameter and transformation applications
8. Applications of likelihood multipliers
9. Survival curve and bivariate estimation applications
10. Nonparametric curve estimation and inference

This book claims to be very general in its approach to curve estimation. Although it is clearly written from the perspective of using Fourier series estimators, the authors do show that the penalized likelihood and kernel methods can also be expressed in this manner. The introduction outlines the philosophical approach from a number of perspectives, and throughout the text, wide use is made of real and simulated data; many figures illustrate the examples. It begins at an easy level, giving a mostly historical perspective, with references to more specialized books. However, later chapters become deeper and more technical. Much of curve estimation is used as an exploratory tool for the purposes of transforming the data and/or for subsequent inference. The authors describe graphical tools, such as the use of the generalized failure-rate function, to explore transformation selection. This book does a great deal to explore and develop the interface between the model-free, non-parametric ap-proach and maximum-likelihood methods as well as some Bayesian inference.

Contents:
1. Preliminaries
2. Same general concepts
3. First-order theory
4. Higher-order thoery: Preliminaries and motivation
5. Some tools of higher-order theory
6. Higher-order theory: Likelihood combinants
7. Higher-order theory: Some further results and tools
8. Various notions of pseudo-likelihood and higher- order theory
9. Further aspects

Readership: Theoretical and applied statisticians with some background in elementary asymptotic theory

This is the inferential counterpart to the authors earlier probabilistic (1989) Asymptotic Techniques for Use in Statistics [Short Book Reviews, Vol. 9, p.25]. It deals with procedures based on various types of likelihood (directed, partial, penalized, profile, pseudo, quasi,...) and, after some introductory chapters, is mainly conerned with higher-order asymptotics both to obtain better approximations and in order to discuss the relative merits of procedures that are equivalent up to first order.
As the author's state, the book "exemplifies concepts and techniques rather than precise mathematical verifications with full attention to regularity conditions". They also point out that much of the field is "undergoing rapid further development" and the account of these parts therefore "has more the flavour of a progress report than an exposition of a largely completed theory".
This report is focused on parametric likelihood theory and thus excludes, for example, asymptotic considerations of nonparametric and semiparametric models, and of robustness.
The book succeeds in bringing together a large amount of useful and interesting work, much of it re-cent and due to the authors themselves. There are close to seventy references to papers by one or both of the authors, most of it done during the last fifteen years.

Contents
1. Introduction
2. Asymptotic inference for (finite-dimensional) parametric models
3. Information bounds for Euclidean parameters in infinite-dimensional models
4. Euclidean parameters: Further examples
5. Information bounds for infinite-dimensional parameters
6. Infinite-dimensional parameters: Further examples
7. Construction of estimates

Many of us knew that this book was on its way and finally it is here: a four-author monograph of more than five hundred pages on the theory of estimation in semi-parametric models. The project started some ten years ago with lectures given by two of the authors (P.J. Bickel and J.A. Wellner) at Johns Hopkins University, Baltimore. The book is on estimation theory based on an independent and identically distributed sample from a probability distribution P, which is thought to belong to a semi-parametric model. That is a collection of distributions which is intermediate between the set of all probability distributions and a usual parametric model. A typical example of a semi-parametric model is a subset which is parameterized by (0,g), where 0 is a Euclidean parameter and g is an element in some function space.
It is nice that the book starts by reviewing estimation theory for parametric models: lower bound theory, geometric results, etc. The next two chapters discuss information bounds for Euclidean parameters in semi-parametric models. This is illustrated by a broad range of examples, including group models, regression models, biased sampling models, mixture models, mis-sing-data models and transformation models. The theory and methods are extended in the next two chapters to the infinite-dimensional parameters in the model. The final chapter deals with methods for construction of estimators and discusses efficiency and other proper-ties.
Readers will certainly be very grateful to the authors for adding a very useful Appendix, of more than 100 pages, with basic definitions and interesting facts from functional analysis, weak convergence theory, etc. Also the very valuable list of more than five hundred references to the literature will be appreciated.
The authors have collected the basic ideas of many famous statisticians of the past and present and successfully brought them together with their own major contributions to the theory. The examples and applicat-ions are very valuable. All this makes "the green book" into a most interesting work for the specialist in the field and for those who want to become specialists.

Contents:
1. Theory of stationary time series
2. ARMA model and model fitting
3. Prediction, filtering and spectral analysis of time series
CASE I : Digital processing of a dynamic marine gravity meter
CASE II : Digital filters design by maximum entropy modelling
CASE III : The spectral analysis of the visual evoked potentials of normal and congenital dull children (Down's disease)
CASE IV : Statistical analysis of VEP and AI by the principal component analysis of time series in frequency domain
CASE V : Periodicity analysis of LH release in isolated pituitary gland by hidden frequency analysis
CASE VI : Statistical detection of Uranian ring signals from the light curve of photoelectric observation
CASE VII : On the forecasting of freight transportation by a new model fitting procedure of time series
CASE VIII: The water flow prediction in Ziang river
CASE IX : Miscellaneous case study

Readership: Undergraduates and postgraduates in time series analysis, applied statisticians, scientists and engineers

The history of time series analysis shows that the subject derives great strength from interacting with other branches of science and technology. This book contains a delightful collection of real case studies conducted in China, using relatively modest computer facilities in the past three decades or so. The collection includes case studies from geophysical exploration, Down's Syndrome, hormone release, Uranian range detection, freight and river flow forecasting, and others. A comprehensive introduction to the theory and method of time series is given prior to the case studies. The book should be a nice compendium to any serious course in time series analysis. The usefulness of the book could be greatly increased if the relevant sets of data were included in the book and the English polished.

Contents:
1. Introduction
2. Stable vector autoregressive processes
3. Estimation of vector autoregressive processes
4. VAR order selection and checking the model adequacy
5. VAR processes with parameter constraints
6. Vector autoregressive moving average processes
7. Estimation of VARMA models
8. Specification and checking and adequacy of VARMA models
9. Fitting finite order VAR models to infinite order processes
10. Systems of dynamic simultaneous equations
11. Nonstationary systems with integrated and cointegrated variables
12. Periodic VAR processes and intervention models
13. State space models

This is a comprehensive review of multiple time series models aimed at graduate students in business and economics. The emphasis is on vector auto-regressive models, though more specific topics such as simultaneous equation models, co-integration, periodic VAR and state space models are also treated. Spectral methods are not covered. The style is quite formal, most statistical properties being derived in theorem-proof format, but the methods are illustrated using a number of real-data examples scattered through the text. The book would be suitable as a text for a grad-uate-level course in statistics or econometrics, aiming at a comprehensive coverage of the theory while also giving the students some insight into the applications.

Contents:
1. Vector time series and model representations
2. Vector ARMA time series models and forecasting
3. Canonical structure of vector ARMA models
4. Initial model building and least squares estimation for vector AR models
5. Maximum likelihood estimation and model checking for vector ARMA models
6. Reduced-rank and nonstationary co-integrated models
7. State-space models, Kalman filtering, and related topics

This book is exclusively devoted to the analysis of multiple-time series. In this the author has tried to provide an exhaustive account of identifiability, estimation, filtering aspects of vector ARMA models, and has illustrated the techniques with real examples. The approach is exclusively time domain, and barely two pages are devoted to spectral analysis. To appreciate the contents of this book one must be familiar with univariate time-series analysis, should have a good knowledge of matrix algebra, and a good know-ledge of multivariate analysis; one definitely must be familiar with canonical correlation analysis. Any reader who requires a thorough account of vector ARMA models will definitely benefit from this book.

Contents:
1. Sums of independent random variables
2. The central limit theorem
3. Convergence of measures, infinite divisibility, and processes with independent increments
4. A celebration of Wiener's measure
5. Conditioning and martingales
6. Some applications of martingale theory
7. Continuous martingales and elementary diffusion theory
8. A little classical potential theory

Readership: Graduate students of probability theory with a good background in probability and modern analysis

The author explains his fascination with interconnections between probability and analysis. Not a first-time basic introduction, this book is valuable for deeper and wider ranging discussions. Its first half deals mainly with sums of independent random variables and with Wiener's measure; conditioning and martingales appear in the second half. Chapter 6 deals with Burkholder's individual ergodic theorem, then develops and uses some theory of singular integrals to help give two derivations of Burkholder's in-equality. Chapter 8 includes discussion of the Dirichlet problem, K.L. Chung's interpretation of the capacitory distribution and Wiener's test for regularity. Numerous exercises to develop ideas further appear throughout the book.

Contents:
PART I : Environmental Monitoring and Sampling
PART II : Measuring Levels and Consequences of Pollution and Contamination
PART III: Climatological and Meteorological Issues
PART IV : Water Resources
PART V : Dynamics of Fish Populations
PART VI : Forestry: Supply and Conservation
Readership: Scientists and statisticians working in all areas of environmental protection and conservation

The eighteen papers grouped into the six parts of this book are the edited proceedings of the SPRUCE Conference held in Portugal during April 1992. SPRUCE is an international initiative concerned with Statistics in Public Resources, Utilities and in Care of the Environment. This book is, on the whole, more coherent and readable than a typical collection of conference proceedings. For this, congratulations can be offered both to the editors, for setting the ground rules and for allowing contributors the necessary space to explain and justify their sometimes complex models, and also to the contributors for supplying their material to a high standard. Statistical aspects of probabilistic models with stochastic and spatial dependence, typically with some non-stationary aspect, give a short summary of the mathematical content of these papers, as grouped under the list of six environmental topics.

Contents:
PART I : Invited Papers
1. Bayesian estimation of fuel economy potential due to technology improvements, R.W. Andrews, J.O. Berger and M.H. Smith
2. Bayes analysis of model-based methods for nonignorable nonresponse in the Harvard Medical Practice Survey, S.L. Crawford, W.G. Johnson and N.M. Laird
3. Use of prior information to estimate costs in a sewerage operation, A. O'Hagan and F.S. Wells
4. Estimation of Bowhead Whale, Balaena mysticetus, population size, A.E. Raftery and J.E. Zeh
5. Bayesian decision support using environmental transport-and-fate models, R.L. Wolpert, L.J. Steinberg and K.H. Reckhow
PART II: Contributed Papers
6. Bayesian analysis of the Ames Salmonella/Microsome assay, R. Etzioni and B.P. Carlin
7. A clinical experiment in bone marrow transplantation: Estimating a percentage point of a quantal response curve, N. Flournoy
8. The composition of a composition: Just the facts, S. Greene and L. Wasserman
9. Predicting coproduct yields in microchip fabrication, W.S. Jewell and S.-K. Chou
10. Synchronicity of whale strandings with phases of the moon, F. Lad and M.W. Brabyn
11. Bayesian predictive inference for small areas for binary variables in the National Health Interview Survey, D. Malec, J. Sedransk and L. Tompkins
12. A cost-utility analysis of alternative strategies in screening for breast cancer, G. Parmigiani and M.S. Kamlet
13. Restoration and segmentation of rail surface images, T.H. Short
14. Assessing mechanisms of neural synaptic activity, M. West and G. Cao

The papers collected in this volume were presented and discussed at a workshop held at Carnegie Mellon University, September 29-October 1, 1991. The volume consists of five major invited papers, each with two discussants and a reply, along with nine contributed papers that were selected from the thirty-four presented at the workshop. Bayesian methods are now far more accessible due to increases in computer power and recent advances in numerical algorithms. The complex models considered in this volume arise from a variety of fields including medicine, business and ecology. The substantial length of the invited articles allows the opportunity for a full justification and discussion of different aspects of the respective models though unfortunately there is a sparsity of graphical representations of models/diagnostics/results. This is disappointing since it is one aspect of the Bayesian approach that is particularly appealing. Overall, how-ever, the volume ably demonstrates the power of the Bayesian paradigm.

Table des matières:
1. Prendre des décisions en univers aléatoire
2. Principe de l'échantillonnage
3. Etude d'une population
4. La statistique du X2 et son utilisation
5. Dépendance et corrélation
6. Outils de repésentation d'un échantillon
7. Représentations d'un échantillon par des cartes
8. Représentations d'un échantillon par des classes
9. La régression linéaire simple
10. Etude de deux populations
11. Etude de plusieurs populations non structurées
12. L'analyse de variance
13. La planification expérimentale
14. De la régression multiple au modèle linéaire
15. Vérification des suppositions
16. Variations sur des relations linéaires
17. Introduction aux modèles non linéaires
18. Observations dépendantes
19. Modélisation de systèmes dynamiques
ANNEXE 1: Calculs, précision et logiciels
ANNEXE 2: Tables statistiques: Programmes et valeurs usuelles

Selon les auteurs, la biométrie est l'ensemble des mathématiques qui sont utilisables dan les sciences de la vie. Comme la hasard joue un rôle important dans beaucoup de ces situations, il est clair que la stat-istique y tient une grande place. Cet ouvrage est des-tinée à un public assez large qui veut apprendre les méthodes principales dans le domaine. Il y a des chapî-tres qui traitent les aspects descriptifs; d'autres sont consacrés à la statistique inférentielle (esti-mation et tests d'hypothèses). La lecture présuppose une certaine connaissance de mathématiques, mais la présentation reste néanmoins ouverte pour un très grand nombre d'étudiants. Les exemples sont nombreux et in-téressants. Dommage qu'il n'y a pas d'exercices.

Contents:
PART I : Communication and Regeneration
1. Heuristics
2. Markov models
3. Transition probabilities
4. Irreducibility
5. Pseudo-atoms
6. Topology and continuity
7. The nonlinear state space model
PART II : Stability Structures
8. Transcience and recurrence
9. Harris and topological recurrence
10. The existence of ð
11. Drift and regularity
12. Invariance and tightness
PART III: Convergence
13. Erodicity
14. f-ergodicity and f-regularity
15. Geometric ergodicity
16. V-Uniform ergodicity
17. Sample paths and limit theorems
18. Positivity
19. Generalized classification criteria
PART IV : Appendices
A. Mud maps
B. Testing for stability
C. A glossary of model assumptions
D. Some mathematical background

Readership: Graduate students from mathematics, engineering, operations research and business

Both theory and applications of Markov chains in discrete time on general state space are discussed. This book nicely adds to the sparsely filled literature gap existing between treatments on Markov chains with countable state space and general theory of Markov processes. The result is a most informative and read-able text on one of the key themes in probability. 'Stability' in the title is to be interpreted in the widest possible sense, i.e. including such topics as recurrence of which many possible definitions exist, ergodicity, general stability for dynamical systems. In a field where one quickly may get lost in often non-standard definitions and a multiplicity of technical results, the authors have managed to convey to the reader a secure path of knowledge and techniques which leads confidently through the entire field. Well-chosen summaries, annotated comments and visually high-light-ing of key definitions and results add considerably to the scientific and pedagogic value of this text. Both researchers and users of Markov chain theory will find this scholarly written book indispensible. A definitive must on one's bookshelf, and an ideal text for use in a post-graduate course on the
subject.

Contents:
1. Preliminaries
2. Convergence of types theorems, symmetry groups and decomposability semigroups
3. Operator-self-decomposable measures
4. Operator-stable measures
Epilogue

In a relatively short period of time nearly sixty years ago, the central limit problem for sums of independent random variables was completely solved: or was it? The classical results, associated with the names of Kolmogorov, Khinchin, Lévy, Gnedenko, Feller among others, impacted greatly the tools, directions and applications of modern probability, particularly through the study and application of the related in-finitely divisible process; for example, Poisson, Brownian, Lévy and Stable processes. But, as Loève quoted in his 1950 review of the central limit theorem (Ann. Math. Statist. 21, 321-338), "No sooner is Proteus caught than he changes his shape."
In rather recent times, attention has been focused upon the more general multi-dimensional form of the central limit problem where one studies the possible limit laws for sums of the form A(X1+X2+...+Xn)-B. Though the form is similar to that of the classical problem, the summands now are vectors and A and B are matrices. It is because of this structure that operator theory, semi-groups and Banach spaces enter so naturally into the discussion.
As in the classical case, many technical analytic details are unavoidable, though it seems from this presentation that there is a certain cleanness to the details that results from this more general approach. The authors provide a fairly complete set of preliminaries in both analysis and probability in their first two chapters, and then provide the solutions to their central limit problems in Chapters 3 (the "class L" or "self-decomposable" case) and 4 (the "stable" case). The book provides a well-written and fairly complete exposition of these recent developments.

Contents:
Glossary of Notations and Abbreviations
1. Introduction, by S.K. Mitra
2. Mathematical foundations of signal processing, by K. Steiglitz
3. Linear time-invariant discrete-time systems, by N.K. Bose
4. Finite impulse response filter design, by T. Saramäki
5. Infinite impulse response digital filter design, by W.E. Higgins and D.C. Munson, Jr.
6. Digital filter implementation considerations, by Y. Neuvo
7. Robust digital filter structures, by P.P. Vaidyanathan
8. Fast DFT and convolution algorithms, by H.V. Sorensen and C.S. Burris
9. Finite arithmetic concepts, by W.K. Jenkins
10. Signal conditioning and interface circuits, by L.E. Larson and G.C. Temes
11. Hardware and architecture, by T. Thong and Y.C. Jenq
12. Software considerations, by J. Fadavi-Ardekan and K. Mondal
13. Special filter design, by P.A. Regalia
14. Multirate signal processing, by R. Ansari and B. Liu
15. Adaptive filtering, by J.M. Cioffi and Y.-S. Byun
16. Special analysis, by R. Kumaresan

Readership: Time-series researchers in statistics, econometrics and the physical sciences

Research in time series analysis is spread amongst the statistical, econometric, electrical engineering and general physical sciences literatures. In electrical engineering much of the research in time-series analysis takes place in the framework of digital signal processing. A main aim of this book is a distillation, primarily from the extensive electrical engineering literature, of the central ideas and primary methods of analysis, design, and implementation of digital signal processing methods. The book itself is huge in both size and coverage, and, with its excellent glossary of notations and abbreviations, is a potentially very useful aid to researchers working at the interfaces of time-series analysis. Each chapter is written by a leading expert, and has its own reference list. The book is very well produced and has a coherent look and feel to it. Of most potential interest to statistical time-series analysts are the chapters on spectral analysis, adaptive filtering (stochastic gradient methods etc.), multirate signal processing (part of the theory of wavelets) and on various fast discrete Fourier transform algorithms with FORTRAN code. I feel that most mathematics or statistics departments could find this book very useful in the increasingly inter-disciplinary world of research.

Contents:
1. Help
2. Science
3. Organised groups
4. The nature of problems
5. Operational research
6. Defining problems
7. Data collection and analysis
8. Models
9. Making choices
10. Implementation and solutions
11. Some organisational aspects of OR
Postscript

Readership: Operational research students, teachers and practitioners, final-year undergraduates and postgraduates

This book covers the art of operational re-search, not its techniques. Mitchell draws on a life-time of experience helping to solve the problems facing organized groups by applying the methods and approaches used by science, the book's definition of operational research. The author emphasizes the needs and aspirations of different clients, singly and together, and the way that a group will have assorted perceptions of the same problems area. The approach of operational research is described in the later chapters. Here there is a clear sense of the messiness and conflicts found in the definition and description of many problems. There are exercises designed to make the reader think; these exercises cannot be lightly solved by turning to techniques on the computer. These and a series of case-studies convey, far better than most books on operational research, the challenges and flavour of the subject.

Contents:
1. The nature of mathematical programming
2. General methods for linear programming
3. Methods for specialist linear programming models
4. Computational implementation of the simplex method
5. Non-calculus methods for non-linear programming
6. General methods for integer programming
7. Computational implementation of the linear programming based branch-and-bound algorithm
8. Specialist methods for integer programming models

In this book the solution methods of mathematical programming are explained by using numerical examples together with a commentary on the character of the methods. The author's aim is to impart under-standing without using a real or pseudo-computer programming language or subjecting the reader to a rigorous mathematical approach. If you are happy with your current textbook on the simplex method, then it is not worth changing; however if you wish a well-written book that in addition has Karmarkar's algorithm, network models, integer programming and discusses computer implementation issues, then I recommend highly this book. This text complements Model Building in Operational Research by the same author; it is suitable for under-graduates and postgraduate students in a technical subject and requires a knowledge of matrix notation.

Contents:
1. Introduction
2. Fundamentals of fuzzy set theory
3. Fuzzy linear programming
4. Fuzzy non-linear programming
5. Interactive multiobjective linear programming with fuzzy parameters
6. Interactive multiobjective non-linear programming with fuzzy parameters
7. Interactive computer programs
8. Some applications
9. Further research directions

The constraints, objectives and data of practical mathematical programming problems do not have the certainty and precision that is assumed by the mathematical theory. This text discusses mathematically this uncertainty and imprecision, and presents interactive algorithms to derive a satisfactory solution for the decision maker. An implementation of these algorithms is available on disk. The programs, which run on a DOS based PC, are limited to seven objectives, forty constraints and forty variables; however, these programs do enable the reader to improve his understanding of the theoretical material. Chapter 8 gives three applications, one linear and two non-linear programs, taken from the author's experience, and illustrates the use of the software on these problems. This book is intended for teachers and students with a background in mathematical programming.

Contents:
1. Manufacturing system design and operation
2. A review of queueing theory
3. Modeling flexible manufacturing systems using queueing networks
4. Modeling production lines using queueing networks
5. Modeling transfer lines and assembly type systems using queueing networks
6. Simulation of manufacturing systems
7. Generative models for the buffer allocation problem in manufacturing systems

Readership: Lecturers, researchers and graduate students in operations research and related areas

Much work has been done on the modeling of manufacturing systems using queueing networks. The authors' aim is to collate these researches and hence promote further research. A background in probability and queueing theory is requisite; advanced mathematics is not. The book begins with two pages of acronyms from mathematics, statistics, operations research, queueing theory, manufacturing systems, and computing, for example, GEM, ETOP, HOCUS, PANACEA. A ten-page glossary of notations follows.
There are two introductory chapters. Each of the next three chapters considers a different type of queueing network models, beginning with a classification of the relevant models according to the under-lying assumptions. Each model is then presented systematically and clearly, with appropriate diagrams and mathematics. The last two chapters examine simulation modeling, a very flexible and popular tool that complements mathematical analysis, and generative models. Every chapter has from two to seven pages of references. An interested and diligent reader will gain much from this reference volume.

Contents:
1. Total quality managementCCnot a quick fix
2. Total quality managementCCmyth or miracle?
3. Understanding total quality management
4. Understanding the change process
5. Establishing the need to change
6. Gaining and sustaining commitment
7. Structure for improvement
8. Problem-solving
9. Using the core concepts as a focus for improvement
10. Education and training for total quality management
11. Quality strategies
12. Making total quality management permanent
13. Implementation of total quality management
14. The future of total quality management

Readership: Anyone seeking a first exposure to the ideas of total quality management (the management, not statistical, issues)

This book is the first in a series of annual supplements to the journal Total Quality Management, entitled Advances in Total Quality Management. The in-tent of the series is for each volume to provide a comprehensive and modern account of its subject. This book would be a suitable primer for newcomers to total quality management, however it does not bring much that is new to researchers or experienced practitioners. Total quality management as a field has matured to the
degree that new books really should provide more extensive cross-referencing of its vast literature. I was disappointed that this book does not deliver it.

Contents:
1. Introduction
2. Sampling
3. Analysis
4. Data Processing
5. Organization

Readership: Analytical chemists and laboratory managers interested in quality control, statisticians lecturing in quality control looking for applications

This is a revised and updated version of a book first published in 1981, and is number sixty in a series of 124 monographs, to date, on analytical chemistry. An extensive range of statistical techniques is shown to be relevant to the achievement of quality control in this particular discipline, with control and cusum charts forming but a small proportion of the whole picture. Techniques described also include t-tests, ranking tests, analysis of variance, sequential analysis, Box-Cox transformations, information theory, Kalman filtering, curve fitting and principal component analysis. Most statisticians will probably find the forty pages of Section 2.6 headed "sampling parameters" to be most novel and interesting. Statistical techniques are given for determining optimum sampling methods from populations of heterogeneous solids, liquids, gases or mixtures rather than from the usual statistical text-book paradigm of an abstract or human population of discrete and separable units.

Contents:
[The 9 'Lives' are case studies].
1. Introduction
The first life : Syringe trouble
2. Decisions and the scientific method
The second life: The pastry man's tale
3. Logic and common sense
The third life: Berwyn Bank
4. Describing a problem
The fourth life: The Happy Hamburger Company
5. Uncertainty
The fifth life: Getting a lift up
6. Deterministic problems
The sixth life: Tattie Fabrix
7. Forecasting
The seventh life: Nirvana residential homes
8. The analyst
The eighth life: Competitive tendering for Conner Mining
9. The anatomy of organisations
The ninth life: Buttermere Oil
10. Bridge building
11. Focal points
12. The analytical process
13. Practical matters
14. The future
15. Closing thoughts
Bibliography: A Journey into Serendipity

Readership: '... aimed at teachers and graduate students in management science, operational research, computing, economics and kindred areas such as accounting and finance ... also ... managers and executives' (Preface); also applied statisticians

The book, written in an attractive style which engages the reader's interest, valuably distils the author's long and wide experience in operational re-search, applied statistics and management. In struc-ture, chapters on various basic aspects of model building for decision making are interleaved with a series of nine case studies which are not necessarily related to the preceding text. Emphasis is placed throughout on the need to understand the realities of the system and to base one's analyses on models which fully relate to this understanding of the real situat-ion. This is clearly illustrated in the case studies. The book is a most welcome addition to the operational research/management science literature.

Contents:
1. Introduction
2. Foundations
3. Generalisations
4. Modelling
5. Inference
6. Remodelling
APPENDIX A. Summary of Basic Formulae
APPENDIX B. Non-Bayesian Theories

Readership: Statisticians and scientists interested in the logic of statistical inference

From the Preface: "This is the first volume of a related series of three - Bayesian Theory, Bayes-ian Computation, and Bayesian Methods - which set out to give an up-to-date overview of our version of the why?, how? and what? of Bayesian statistics. The second volume will be written by Alan E. Gelfand and Adrian Smith; the third volume will be written by the current authors. The original motivation for this enterprise stemmed from the impact of de Finetti's two-volume Theory of Probability ... .
This volume on Bayesian Theory sets out to provide a fairly complete and up-to-date overview of what we regard as the key concepts, results and issues."
The authors succeed in their ambitious aims to a remarkable extent. I know of no other work which explores the case for Bayesian methods in more useful detail. Their account of what they call "frequentist methods" is less thorough. But the Users of "frequentist methods" may note the absence of "experimental design", "known facts", and "established scientific theories" from the listed sources of models in Chapter 4.

Contents:
1. Coins, dice and cards
2. Conditional probability
3. Distributions
4. Expected value
5. Limit theorems

The treatment is a classical, non-measure-theoretic approach to undergraduate level probability. The particular niche market targeted by this text is Math 471 at Cornell University. Introductions to probability that cover essentially the same subject matter and which motivate the material at least as well are widely available. It is not obvious to this reviewer that a wider audience can have been intended. That said, explanations are clear and there are plenty of exercises.

Contents:
1. A historical survey
2. Properties of the inverse Gaussian distribution
3. Characterizations
4. Combinations, extensions, and relatives
5. Inverse natural exponential families on R
6. Statistical properties

Readership: Those interested in exponential families, the inverse Gaussian distribution and probability distributions in general

This book is an important addition to the literature on the inverse Gaussian distribution. It begins with an excellent chapter on the origins of the distribution. The results for the inverse Gaussian dis-tribution are developed as a special case of more general exponential families and there is an important chapter on inverse natural families. There is a concise presentation of sampling results for statistical inference, but the emphasis of the book is on the probability distribution.

Contents:
1. Frequency distributions
2. Measures of location and dispersion
3. Moments and cumulants
4. Characteristic functions
5. Standard distributions
6. Systems of distributions
7. Multivariate distributions
8. Probability and statistical inference
9. Random sampling
10. Standard errors
11. Exact sampling distributions
12. Cumulants of sampling distributions (1)
13. Cumulants of sampling distributions (2)
14. Order-statistics
15. The multinormal distribution and quadratic forms
16. Distributions associated with the normal

In 1943, The Advanced Theory of Statistics, Volume 1, by M.G. Kendall appeared, Volume 2 appearing in 1946. The first volume had five editions, the last appearing in 1952, Volume 2 had three editions, the last appearing in 1951 with second and third impressions. In 1958, the work was started again with three volumes, the first subtitled `Distribution Theory' and written jointly with A. Stuart. Three subsequent editions appeared in 1963, 1969 and 1977.
A third author, J.K. Ord, was taken on board for the fifth edition. Sir Maurice Kendall passed away in 1983. His association will remain and remind future generations of statisticians of the amount of work he put into this with the new title Kendall's Advanced Theory of Statistics. The fifth edition of Volume 1 was published in 1987 [Short Book Reviews, Vol. 7, p.23.].
This, the sixth edition, is again a thorough revision of the previous one. The book is overall larger by almost eighty pages including those for references which has been increased by eight pages. A new Chapter 7 treats bivariate and multivariate distributions. Chapter 16 covers multivariate sampling theory which was previously in Volume 3. New material includes skewness and kurtosis, hazard rate distributions, the bootstrap, the evaluation of multivariate normal integral ratios of quadratic forms, forty new exercises and twenty new examples. Statisticians should thank the authors for continuing to keep student, teacher and researcher up-to-date.

Contents:
A brief introductory guide to S-PLUS
1. Data description and simple inference: IQ scores of children of depressed and non-depressed women
2. Predicting the volume of black cherry trees from measurements of their height and diameter
3. Analysis of variance: Diets for chickens
4. Logistic regression: Predicting outcome to treatment for patients with Leukaemia
5. Survival analysis: Modelling the survival time of patients with Leukaemia
6. Nonlinear modelling: Fitting the Michaelis-Menten equation to hormone-receptor assay results
7. The analysis of time series: Yearly numbers of sunspots
8. Principal components: Exploring the progress of competitors in a 100-kilometre road race
9. Cluster analysis: Classifying countries in terms of the athletic prowess of their women
10. Correspondence analysis: Suicide behaviour in the former West Germany
11. Scattergrams and density estimation: birth and death rates for 69 countries

S-PLUS is a commercial implementation of the S language (Becker, Chambers and Wilks, 1988 [Short Book Reviews, Vol. 9, p.2] and Chambers and Hastie, 1992 [Short Book Reviews, Vol. 12, p.4]). This handbook uses eleven case studies to give an overview of how to use it. Each case study covers a set of data, a capsule description of some statistical methods, and then shows how to implement them in S-PLUS. A few exercises are given at the end of each chapter. There are about sixty-five figures and thirty-three tables, including listings of all of the data. Programming in S-PLUS is not covered.
The sets of data are taken mostly from other texts or the statistical literature. Some of the analyses are weak, for example, the nonlinear regres-sion chapter does not attempt to assess the errors in the parameter estimates, but this is forgivable, given the size and aim of the book. The S-PLUS code is pre-sented clearly and is well-annotated.
A minor inconvenience is that the book was written for S-PLUS for DOS, apparently version 3.0, which is no longer in production. Almost all of the code in the book will work equally well in Unix or Microsoft Windows; a single page would have sufficed to describe the differences.
This is a very useful handbook. It is too small to stand on its own, but its concentrated struc-ture makes it a far more accessible introduction and quick reference to S-PLUS than the over 3000 pages of S-PLUS manuals and S books.

Contents:
1. Introduction
2. Data collection and graphical summaries
3. Numerical data summaries: Location and dispersion
4. Probability and distribution
5. Sampling, estimation and confidence
6. Simple tests of hypotheses: `Significance tests'
7. Control charts for process management and improvement
8. Control charts for average and variation
9. Control charts for `single-valued' observations
10. Control charts for attributes and events
11. Control charts: Problems and special cases
12. Cusum methods
13. Process capability: Attributes, events and normally distributed data
14. Capability: Non-normal distributions
15. Evaluating the precision of a measurement system(gauge capability)
16. Getting more from control chart data
17. SPC in `non-product' applications

Readership: Industrial practitioners needing a reference text for statistical methods useful in quality control and improvement

The democratization of "quality" in manu-facturing during the past decade has yielded impressive gains by large numbers of workers armed with simple tools like Pareto charts, histograms, and basic control charts. This book is for the smaller number of practi-tioners who have outgrown the basic tools and are ready for a step towards statistical sophistication. Although the first six chapters cover very traditional statisti-cal topics, the remaining chapters on control charts, process capability, and gauge capability are useful expositions that go into more detail than the majority of industrial texts I've recently encountered. The book is not a primer. It assumes a relatively high level of numeracy, although calculus is not required. The ment-ion of "TQM" ("Total Quality Management") in the book's title is almost incidental. TQM provides a con-text for data collection and statistical analysis, but TQM itself is not addressed in depth.

Contents:
1. Introduction to statistical quality control
2. Frequency distributions
3. Area under normal curve
4. Probability concepts
5. Process control: X-bar and R control charts
6. Analysis and decision making using x-bar and R control charts
7. Attributes control charts
8. Other types of control charts
9. Operating characteristic curves
10. Designing acceptance sampling plans with specified producer's and consumer's risk
11. Dodge-Romig rectifying inspection acceptance sampling plans

This is a practically oriented recipe book for the basic techniques of statistical quality control. The text is built around demonstrating these techniques on realistic looking data sets, with all the relevant calculations set out in full detail. The discussion questions and problems at the end of each chapter, plus the supplied computer program, would seem to make this a fully self-contained course for the targetted audience. The book could also be used as a source of data and problems for teachers of more mathematically oriented courses. Such teachers might also be interested to read how this book manages to have a chapter on the normal distribution before the chapter on probability.

Contents:
PART I : Pattern Algebra
PART II : Pattern Topology
PART III: Pattern Dynamics
PART IV : Metric Pattern Theory
PART V : Pattern Deformations
PART VII: Creating Regular Structures
Postscript

Readership: Mathematicians, computer scientists, statisticians and other scientists

The 'Advice to the reader' section of this book begins by saying: 'Reading this book will require a determined effort.' Given that the book is two inches thick and weighs over two kilograms, this might be seen as superfluous!
The aim of this book is to present a mathematical foundation for representing and reasoning about patterns and their variability in the broadest sense. The book is the culmination of three decades of work by the author.
The volume opens with a description of the mathematical terms and concepts involved and continues in the definition, theorem, proof, style of a mathematics text. It formalizes the concept of regularity in terms of atomic particles (generators) comprising a pattern, their relationships (connectors) to other generators, and the properties of the links (bonds) between them. These formal descriptions are then supplemented by 'rules of interpretation', expressing the relationship between abstract configurations and ob-served patterns. Later on, these regular structures are embedded in spaces upon which probability measures are defined. The penultimate part of the book describes how the ideas are applied in inference about patterns, in areas such as image restoration and pattern recognition.
Stress is placed on practical work as well as mathematical analysis, using APL to simulate regular structures so that one can learn about them.
I would hazard the assertion that most modern science is analytic in structure, but that many of the big steps typically arise from a new synthesis. This book is just that, an impressive new synthesis. It clearly represents the bible of pattern theory, and from that perspective one cannot but recommend it. Having said that, a one-hundred-page introductory monograph outlining the theory would also be valuable to a wider scientific community.

Contents:
1. The processes of science
2. Principles of experimental design
3. Survey of experimental design: A preview
4. Linear model theory
5. Randomization
6. The completely randomized design
7. Comparisons of treatments
8. Use of supplementary information
9. Randomized block designs
10. Latin square type designs
11. Factorial experiments: Basic ideas
12. Response surface designs
13. Split-plot type designs
Additional reading from the writings of O. Kempthorne

In 1952, O. Kempthorne published a major book which for many years was the most systematic account of what may be loosely called the traditional theory of experimental design. Regrettably this book was al-lowed to fall out of print. The present book, the first of two volumes, is based to an appreciable extent on the earlier one, but is much more than a second revised edition.
It begins with a thirty-page critical setting of philosophical background, the role of theory and experiment in science, the nature of causality and so on. There is a quite thorough account of the algebra of the linear model. The role of randomization in determining the model and error structure appropriate in any particular experiment is described with con-siderable care. Of the key concepts of the subject, error control and elaborated treatment structure, the first gets major emphasis in the first three hundred and fifty pages. The study of factorial experiments emphasizes, although not exclusively, the two-level case; rather surprisingly response surface designs are treated as a largely separate topic and the book concludes with a strong account of split-plot experiments and a brief, and not too satisfactory, few remarks on so-called repeated-measures designs regarded more or less as an appendage to split-plot experiments. Nine of the thirteen chapters have exercises.
This too is a major book. All those with a serious interest in the subject, whether as research workers in the field, as applied statisticians using the ideas or as teachers will profit from it.

Contents:
1. Basic principles
2. The field plot
3. Basic experimental designs
4. Agronomy trials
5. Variety trials
6. Combined analysis of several experiments
7. Experiments with perennial crops
8. Pasture trials
9. On-farm trials in farming systems research
10. Intercropping research
Readership: Agricultural field researchers

The author covers all major topics where an agricultural research worker should use statistical techniques, discussing them and interpreting the results clearly as to a colleague; there is a feeling almost that the author is present. The worked examples, of which there are many, are more than illustrations, for example, covariance is introduced while discussing an example in perennial crops. Computer software is ignored, which avoids some complications, but is limiting; inter alia, modern methods of design and analysis requiring computer help, for example, á-designs and neighbour methods, are not mentioned. There is an at-tempt to treat controversial topics from the practical point of view; recent controversy on intercropping is discussed, but only Land Equivalent Ratios are exemplified; recent controversy in genotype stability is ignored; it is unfortunate that an outmoded analysis is recommended. There are some typographical errors in the data and text, but this is a well-thought-out book by an experienced applied statistician, and should be of immense help to all agricultural researchers.

Contents:
1. The nature of agricultural research
2. Key assumptions of experimental designs
3. Designs for reducing error
4. Single-factor experimental designs
5. Two-factor experimental designs
6. Three (or more)-factor experimental designs
7. Treatment means comparisons
8. Sample design over time
9. Regression and correlation analysis
Readership: Agricultural and biological researchers

This book is an introduction to the basic concepts of experimental design and analysis, with a very minimum amount of mathematical explanation. There is a useful introduction, and a wide range of practical examples, many of which are accompanied by details of the appropriate calculations. The author claims that no previous knowledge of analysis of variance is needed, but I think that researchers with relatively little training in statistics would find the book hard to follow, especially as the material tends to dart backwards and forwards between topics.
I wonder why block designs, other than balanced and partially balanced lattices, are not introduced. This would remove the restriction that the number of treatments be a perfect square. There are numerous errors in the book, of which I shall mention one. In Chapter 9, the formula for the least-squares estimate of the slope in a simple linear regression model has a faulty denominator. Readers might find it confusing that no distinction is made between unknown parameters and their estimators.

Contents:
1. Introduction
2. Design considerations
3. Exploring longitudinal data
4. General linear models for longitudinal data
5. Parametric models for covariance structure
6. Analysis of variance methods
7. Generalized linear models for longitudinal data
8. Marginal models
9. Random effects models
10. Transition models
11. Missing values in longitudinal data

Readership: Final year undergraduate or first year postgraduate students of statistics

Recent years have witnessed the publication of a number of books on the analysis of repeated measures and longitudinal data. Like some of the others, this one attempts to present a unified description of the range of approaches to analyzing such data. So how does this one differ from the others? It explicitly distinguishes between three classes of models: marginal models (in which forms for the expectation and variance are separately modelled), random-effects models (in which separate regression models are fitted to each subject, with the coefficients randomly selected), and transition models (in which a model for the current response includes previous responses amongst the covariates). It includes a chapter discussing the contrasts between the three classes. It also explicitly distinguishes between two types of problem: those where the regression of the response on the covariates is of primary interest and those where the correlation structure is of primary interest. The book discusses continuous, binary, and count responses and also has a chapter on design considerations, including issues such as sample size, something which is particularly welcome.
Inevitably one would have done some things differently oneself. For example, in some disciplines the multivariate-analysis-of-variance method is very heavily used, a consequence of the availability of routines such as the SPSS multivariate analysis of variance. Although this is implicit in much of the discussion, it would have been nice to see it described explicitly: multivariate analysis of variance does not even feature in the index. None-the-less, I enjoyed this book greatly. It is exceptionally lucid, explain-ing the mathematical detail clearly and accurately. I thoroughly recommend it.

Contents:
1. Introduction
2. Matrix theory
3. Multivariate normal and related distributions
4. Introduction to linear models
5. Parameter estimation
6. Statistical inferences
7. Linear regression models
8. Analysis of variance models
9. Analysis of covariance models
10. Variance components models

Readership: Upper-level undergraduate and graduate students in statistics or related areas

This is an advanced unified textbook on linear models. Its technical level is slightly higher than Seber's Linear Regression Analysis. About sixty pages are devoted to matrices; the elementary definitions, and the ideas of space, dimension, and rank are assumed known. The treatment of topics is quite comprehensive. The emphasis is mainly theoretical; about a dozen small sets of data are used as illustrations, most of these being in Chapters 7-10. Nearly all of the one hundred and eight exercises are on theory. The dates of the references go up to 1990 with the exception of a few later references written by the authors themselves.

Contents:
PART I : Logistic Regression With Two Categorical Covariates
PART II: Generalizations
Readership: Applied statisticians, graduate students in biostatistics and epidemiology

A basic concern in many scientific investigations is the assessment of the simultaneous influence of several factors on a quantity of interest. The response may be binary and not all covariates may be measured for all units. The assumption of missing at random, which excludes a dependency between the true covariate value of a potentially unobserved covariate and its observability, has been fundamental to statistical methods handling missing values.
Two stronger restrictions than the missing at random assumption are introduced in this text for logistic regression with two categorical covariates and missing values in only one covariate. Censored, grouped and miss-measured covariate values are not considered, merely the situation where the value of a covariate is completely unknown. Basic difficulties with the generalization for more than two covariates with arbitrary patterns of missing values are described but many questions remain unanswered.
At times, the notation is cumbersome and the subject index is severely limited. However, the text is worth reading especially if one is contemplating working in this important area.

Contents:
1. Introduction
2. Interpolating and smoothing splines
3. One-dimensional case: Further topics
4. Partial splines
5. Generalised linear models
6. Extending the model
7. Thin plate splines
8. Available software

Smoothing and interpolation are two of the oldest activities in applied statistics, and there is a wide variety of methods for carrying them out. In the past fifteen to twenty years there has been much research in non-parametric methods, with terms like splines and penalized likelihood or penalized least-squares being very much in the air. The present volume, written by two active workers in the field, provides an excellent introduction to these and related tech-niques. It explains and motivates the various methods and results, and includes some of the proofs, while never becoming bogged down in the details.

Contents:
1. A wide range of applications
2. A simple test
3. Testing hypotheses
4. Experimental designs
5. Multivariate analysis
6. Categorical data
7. Dependence
8. Clustering in time and space
9. Coping with disaster
10. Which statistic? Solving the unsolvable
11. Which test should you use?
12. Publishing your results
13. Increasing computational efficiency
14. Theory of permutation tests

This book has an excellent bibliography, in four parts, that takes up about forty pages out of the total. The author has an interesting, cook-book style which adds to the readability of the book. Many modern, as well as classical techniques are included; thus the book appears to be up-to-date. Although there are questions at the end of many chapters, they are not the kind of exercises that one would expect in an introductory text on this topic. Thus in an introductory book one would expect to find exercises for the student to practice on and answers at the end of the book in order that he can check his results. There are no such questions or answers in this book. In the first solid ex-ample on page 30, the author adds the ranks 1, 2, 3, 4 and 10 and gets 15 instead of 20. Furthermore the count of 24 in the next paragraph is left as an exercise. I only got 19; I may be wrong by 1 but not by 5. What a pity to spoil all that good work by not checking for miscalculations and/or misprints. On page 25 the author rejects the hypothesis 0?00=0 at level 1/16 with the sample - 1,2,3,1.1,5, but on page 27 with the same data and the same level he accepts the same hypothesis. I believe the latter is the result of poor wording.
It is almost impossible to use this book as a basis for learning the fundamentals of the theory of permutation tests. However, the book serves a useful
need for quickly finding different papers on the appliation of resampling methods which includes permutation tests.

Contents:
1. Introductory concepts
2. Assumptions and modelling philosophy
3. Statistical theory
4. Line transects
5. Point transects
6. Extensions and related work
7. Study design and field methods
8. Illustrative examples

The estimation of the size of a biological population, or its density (number of individuals per unit area) in a given region is conceptually straight-forward if all individuals can be counted within designated area units, known as quadrats. The problem becomes much more interesting if, as in the case with most animal populations, detection of an individual by the observer is uncertain. The related methods of line transect and point transect sampling have been developed to deal with this situation, and each of the authors of this book has played a prominent part. The scope of the book reflects the breadth of expertise amongst the authors, covering statistical theory and method, survey design, field sampling technique and software. The result is a comprehensive, lucid account which can be highly recommended to anybody who wishes to learn about this relatively specialized topic.

Contents:
1. Background on probability and statistics
2. Stochastic differential equations
3. Introduction to discrete time approximation
4. Strong approximation
5. Weak approximation
6. Applications

Readership: Students in mathematics, physics, engineering and economics
This is a computer aided introduction to the numerical solution of SDE using PC experiments. A floppy disk containing Borland TURBO PASCAL programs (only IBM compatible MS-DOS PCs) for over 100 problems is provided together with comments on the programs and some hints on their use. This book is intended for readers without specialist stochastic background but is related to the more theoretical monograph Numerical Solution of Stochastic Differential Equations, by P.E. Kloeden and E. Platen. [Short Book Reviews, Vol. 13, p.7.]

Contents:
1. Introduction
2. Classification
3. Classical statistical methods
4. Modern statistical techniques
5. Machine learning of rules and trees
6. Neural networks
7. Methods for comparison
8. Review of previous empirical comparisons
9. Dataset descriptions and results
10. Analysis of results
11. Conclusions
12. Knowledge representation
13. Learning to control dynamic systems

This multi-authored book arose from the European Community (ESPRIT) project StatLog which compared and evaluated a range of classification techniques on a number of large, real sets of data. The first six chapters provide readable introductions to all the methods considered; the next two discuss general issues of method comparison; the following three chapters describe the data, analyze the results and draw conclusions; and the final two chapters deal with knowledge representation and dynamic control theory.
There is much of interest here, spanning an introduction to classification, a cross-disciplinary comparison of methodology and an empirical evaluation of performance. The editors have done a good job of integrating the different contributions into a unified whole, and relatively few glitches remain. The appendices give valuable information on sources of data and statistical algorithms. In summary, a book, which succeeds in contributing to cross-fertilization between machine learning, statistics and neural networks, is one which should be of interest to all workers in classification.

Contents:
1. Crisp and fuzzy sets in statistics
2. The likelihood formulation of the fuzzy set partition
3. Estimation of the parameters of the GoM model
4. A GoM model for aggregate data
5. Longitudinal and event history forms of the GoM model
6. Empirical Bayesian generalizations of the GoM model
7. Forecasting and simulation with fuzzy set models
8. Fuzzy set analyses of combined data sets: a model for evaluation studies
9. Area of further statistical research on fuzzy sets

This book is mainly concerned with the model-ling of high-dimensional, sparse, categorical data. It describes an interesting new computer-intensive statistical methodology based on fuzzy sets as opposed to classical discrete (crisp) sets. Each element in a
fuzzy set has a grade of membership (GoM) score; elements can have partial membership of multiple fuzzy sets. Fuzzy partitions of collections of elements can be set up which produce a set of GoM scores for each element relative to 'extreme profile' sets. These properties are exploited to describe non-stochastic heterogeneity within the authors' GoM models. The underlying mathematical theory is given and the statistical properties of the GoM models are developed. These models can be viewed as generalizations of standard multivariate models. The authors compare the new and standard models both theoretically and empirically on a variety of data sets. The new models appear to per-form better than the standard models.

Contents:
0. Conventions and notation
1. Operations on sets
2. Classes of subsets of a space
3. Set functions
4. Measure spaces
5. Measurable functions
6. Integration
7. Hilbert space
8. Convergence of measure sequences
9. Signed measures
10. Measures and functions of bounded variation on R
11. Conditional expectations; Martingale theory

This book is described in the introduction as a work not originally planned to be published, but as a chance for the author to organize his own ideas. Would that more books had such beginnings. It provides an extremely clear introduction to the subject in some depth, and is enjoyable to read. One of the author's ideas is that probability is an essential part of measure theory. This adds obvious richness and interest to what is often seen as a rather dry subject. On a more technical level, another novelty is the focus on pseudo metric spaces as a means for dealing with the usual nuisance of pretending that functions which are the same almost everywhere are actually identical. (A pseudo-metric space allows non-identical elements to be at zero distance from each other.) The lack of exercises may, perhaps unfairly, limit the book's appeal as the sole accompaniment to a graduate measure theory course.

Contents:
1. Introduction
2. Hilbert spaces
3. Probability theory
4. Estimating functions
5. Orthogonality and nuisance parameters
6. Martingale estimating functions and projected likelihood
7. Stochastic integration and product integrals
8. Estimating functions and the product integral likelihood for continuous time stochastic processes
9. Hilbert spaces and spline density estimation

The authors introduce the theory of probability by making the inner product on a Hilbert space the basic concept. This is in contrast to the more common Kolmogorov approach with probability measures on a sample space. Random variables appear as elements of a Hilbert space and product moments between random variables are given by the inner product.
The major theme of the book is to show that the Hilbert space methodology is essential to the development of probability and statistical inference. One of the great advantages is that many common analytical tools obtain a geometrical interpretation in the Hilbert space approach. The concepts of conditional expectation and various projections are nice examples. Projections, subspaces, orthogonal decompositions, ... from Hilbert space theory become the basic machinery for explaining concepts like sufficiency, likelihood, stochastic integration, ... .
To enjoy this book it is essential that the reader is already familiar with linear algebra and the classical theory of probability, statistics and stochastic processes. The reader will certainly also appreciate the notes at the end of each chapter giving some interesting historical background.

Contents:
VOLUME I
1. Measure theory
2. Probability measures
3. Integration
4. Expectations and moments
5. Convergence in law
6. Conditional expectations
7. Martingales
VOLUME II:
8. Random vectors and their densities
9. Stochastic processes
10. Regular conditional probabilities
11. Optimal stopping strategies
12. Exponential families
13. Consistency of maximal estimators
14. Asymptotic normality

Readership: Instructors, graduate students and researchers in probability and mathematical statistics

These two substantial and significant volumes, consisting of fourteen chapters, four hundred and eighty-four sections and one thousand one hundred and seventy-four pages, represent an exceptionally thorough and scholarly treatise on modern probability and selected major topics in mathematical statistics. The volumes provide an exciting and potentially profitable opportunity, both for reading and teaching. They come forty years after the first modern advanced probability text by M. Loève, the influence of which can still be seen in these volumes. The books are comprised of many short sections devoted to single topics. They make the material's development easy to follow and to reference. Everything beyond undergraduate algebra and analysis is included, making this systematic presentation
reasonably self-contained. The reader who masters the material in these volumes will be well qualified indeed to understand and use modern probability, and to study advanced topics in stochastic processes and mathematical statistics. The coverage of probability is extensive; that of statistics is selective. This is consistent with the title's description of "a view toward statistics" provided the view is understood to be from a vantage-point overlooking a particular subregion of this broad discipline.
The author emphasizes his inclusion of historical backgrounds; this is well done, especially for the earlier stages of probability. Recent history is not as well honoured; more names, dates or references for major results of this century would be helpful. Each chapter includes a large collection of exercises (376 in Volume I and 160 in Volume II), mostly very challenging and instructive. Many have hints and many are major results that in other books might be included in the main text; (for example the plague, eelpouts, a fish auction, gambling - with Molly and Dick - and risk probabilities) as well as some interesting in-sights into more familiar examples. In several places the author uses examples to urge the reader to rely on computations rather than intuition. The apparently contradictory advice on page I-474 is due to a typographical error or is it? After all, it is a probabilists unique intuition that has led to many of the recent successes of the probabilistic methods in analysis. Also, as the moral (not "morale") of an historical aside about M. Souslin, the reader is admonished in I.1 to "Never trust your textbooks too much!". Fortunately, the theoretical substance of these two books seems to be exceptionally trustworthy.
The full enjoyment however, of the core material in these volumes has been significantly marred by the presence of an excessive number of typographical errors, for example there are 13,8 and 8 slips on pages I.xxi, I.317, and II.470, respectively, and 27 in the combined prose of reference sections 1.51, 8.46 and 12.31, which comprise a total of only 54 lines. Most of these slips are grammatical errors of tense and declension; others involve extra or omitted words. Though a few misspellings do occur including names, for ex-ample, Lehman, and omitted accents in Loève and Lévy, practically all of the errors involve the incorrect use of correctly spelled words. This indicates good use of computer spelling checks. However, the annoyingly large number of writing errors that appear in the longer prose sections, and especially in the reference sections to the chapters, indicates insufficient editing efforts of the basic type that long preceded computers. The errors are so pervasive that the publishers should stop further sales and recall copies already sold. Even though English is not the author's first language, the prose within the main theoretical parts is well writ-ten. Thus one assumes that the unacceptably large number of typographical errors in other portions of the material is due to some last minute haste in the publishing process. I do not fault the author primarily in this matter. Although he could undoubtedly have arranged directly for additional editing, the publishers have a basic responsibility, with the help of their academic editors to ensure that the quality of their products attain at least minimal levels of read-ability and presentation. More subjective in nature is the complete lack of punctuation at the end of displayed expressions; the regular use of contractions don't, aren't, didn't, couldn't, wasn't, etc.; and the lightness of the type. Moreover, the first eight pages of the Preface to my copy of Volume I separated from the binding shortly after I began to read.
Setting aside my disappointment in the low level of editorial effort expended upon these books, I conclude by re-emphasizing the high quality of the
books' basic contents. Because of this, these two volumes indeed provide a fresh and exciting opportunity for researchers, instructors and students alike. This was a difficult review to write, not because of the need to comment on editorial short comings, but because the high quality of the main material made it difficult for me to put the books down long enough to complete this review.

Contents:
1. Introduction
2. Conditional probability and conditional expectations
3. Markov chains: Introduction
4. The long run behavior of Markov chains
5. Poisson processes
6. Continuous time Markov chains
7. Renewal phenomena
8. Branching processes and population growth
9. Queueing systems

Readership: Advanced undergraduates and graduate students in probability and statistics

Except for a complete reprinting, this is largely the same book as the first (1984) edition [Short Book Reviews, Vol. 5, p.45]. The main change is a doubling in the number of exercises. The book is an excellent introductory treatment of stochastic processes, at a somewhat lower level than the same authors' First Course in Stochastic Processes. The main distinctive feature of the book is the discussion of de-tailed applications as diverse as crack growth in materials, the strength of bundles of filaments, simple genetic models, plasmic reproduction, insect control, redundancy and the burn-in phenomenon, and queueing networks. It will form a suitable text for an advanced undergraduate or beginning graduate course focusing on applications of stochastic processes.

Contents:
1. Preliminaries
2. Computation of expectations in finite dimensions
3. Simulation of random processes
4. Deterministic resolution of some Markovian problems
5. Stochastic differential equations and Brownian functionals

Readership: Students and researchers in probability, applied probability, engineering and computer science

The area described by the title of this book is young and has by no means reached a stage where its contents are well-defined or a huge variety of methods have been developed. The authors do not claim to be comprehensive but take up some of the possible themes, for example, properties of random number streams, the Monte Carlo methods, simulatability, simulation of stationary fields and processes, numerical balayage, Lévy processes and discretization methods for SDE's. Some of the topics which belong to the area, but are not discussed, are variance reduction in simulation, steady-state simulation, rare events simulation, trans-form inversion, linear algebra methods for large Markov chains and Neuts' matrix-algorithmic approach to queue-ing theory. The style of the book is quite abstract, stressing the mathematical foundations rather than the computer implementation. It is a welcome contribution to an important field still in development.

Contents:
1. Preliminary remarks
2. Brownian motion, Poisson processes, á-stable Lévy motion
3. Computer simulation of á-stable random variables
4. Stochastic integration
5. Spectral representations of stationary processes
6. Computer approximations of continuous time processes
7. Examples of á-stable stochastic modeling
8. Convergence of approximate methods
9. Chaotic behaviour of stationary processes
10. Hierarchy of chaos for stable and ID stationary processes

Readership: Mathematicians, probabilists, statisticians, computer scientists, experimental scientists, economists

This monograph is about Lévy á-stable processes which play an important role in a wide range of problems of stochastic modeling of random phenomena. Using statistical estimation techniques, computer simulation and visualization procedures, and numerical discretization methods, the authors succeed in producing a remarkably impressive and very useful approximate numerical technology for studying these stochastic processes. The intended readers of this book are assumed to be familiar with basic concepts from stochastic integration, stochastic differential equations, convergence of approximate methods, statistical estimation, numerical discretization, and some more, which are introduced without proofs. It may, however, be followed by less specialized readers as well, for example by those interested mainly in computer simulations and graphic representations. Most of the results, including the two original topics of this monograph: computer simulation of á-stable processes and their chaotic behaviour, are presented with detailed proofs. In the Appendix the computer program STOCH-Lm.c is presented. It was employed by the authors to produce many graphical examples in this book. Running this program, one can approximately solve stochastic differential equations with respect to the á-stable Lévy motion.

Contents:
PART I : Stochastic Models and Their Forecasting
PART II : Stochastic Model Building
PART III: Transfer Function Model Building
PART IV : Design of Discrete Control Schemes
PART V : Charts and Tables
PART VI : Exercises and Problems

Readership: Statisticians with an applied interest in time series, econometricians

This book is essentially a reprinting of the well-known book by Box and Jenkins which first appeared in 1970. A new chapter, Chapter 12, has been added on intervention analysis. The basic idea here is that some special external event occurs at a known time T. This is treated by assuming the presence of a pulse or step input at time T and then applying traditional methods of transfer function estimation. Also, some changes have been made to the other chapters, although this has left the central themes unaltered.
The book remains a useful practitioner's guide to univariate time series analysis using ARIMA type models. The book provides excellent motivation for these problems and describes some strategies which have been found to be useful in practical problems. Thus, the book continues to be a useful reference source for applied time series analysts.

Contents:
1. Environmetrics, science and decision making
2. Basic statistical concepts
3. Stationary nonseasonal models
4. Nonstationary nonseasonal models
5. Model identification
6. Parameter estimation
7. Diagnostic checking
8. Forecasting with nonseasonal models
9. Simulating with nonseasonal models
10. The Hurst phenomenon and fractional Gaussian noise
11. Fractional autoregressive moving average models
12. Seasonal autoregressive integrated moving average models
13. Deseasonalised models
14. Periodic models
15. Forecasting with seasonal models
16. Causality
17. Constructing transfer function-noise models
18. Forecasting with transfer function-noise models
19. Building intervention models
20. General multivariate autoregressive moving average models
21. Contemporaneous autoregressive-moving average models
22. Exploratory data analysis and intervention modelling in confirmatory data analysis
23. Nonparametric tests for trend detection
24. Regression analysis and trend assessment
Readership: Environmental scientists, water resource engineers, hydrologists

The authors use examples from the two fields of statistical water quality modelling and stochastic hydrology to illustrate methods of environmetrics. An important section is devoted to time-series theory which is dealt with both clearly and rigorously using Box and Jenkins methods, but also recent developments in the field. After presenting models for stationary and nonseasonal models, nonstationarity, seasonal,
deseasonalized and periodic models are examined separately. Essential tools for environmental impact assessment, transfer function-noise and interventions models and their wide applicability to the many variables of hydrological systems are then presented extensively.
This book fulfils a clear need in the field of water resources and more generally provides a de-tailed review of methods for statistical analysis of environmental systems. The section on long-memory modelling is particularly useful and a welcome emphasis is given to non-parametric tests in confirmatory data analysis.

Contents:
PART 1: Introductory Chapters
PART 2: Multirate Filter Banks
PART 3: Special Topics
PART 4: Multivariable and Lossless Systems
Readership: Research students and researchers in electrical engineering and statistics

One of the strands of theory and practice from which the coherent study of wavelets has emerged is that of multirate systems and filter banks. These filter banks consist of components which increase or de-crease sampling rates, often by a factor of 2, the so-called dyadic case, combined with low-pass and high-pass filters. Filter banks perhaps provide the most appealing route into wavelet theory for the time series and spectral analysis researcher. This book is written by one of the most respected researchers in this area, who also writes in a very clear and lucid style. Whilst the book is obviously directed primarily at electrical engineers, many of the key concepts in the filtering view of wavelets, such as perfect reconstruction, quadrature mirror filters, paraunitary filter banks etc. are well developed in this extremely well-written book, and hence available to a wider audience. The book can thus act as a very valuable resource for those wishing to follow past and future developments of wavelet theory.

Contents:
1. The problem of output regulation
2. Disturbance attenuation via H? methods
3. Full information regulators
4. Nonlinear observers
5. Nonlinear H? techniques

This book is concerned with the design of a feedback regulator for a system such that its output is caused to track a desired reference signal with zero, or at least small, error. The system is assumed to have two inputs; one which can be manipulated and one which captures the influence of external disturbances. The reference trajectory and disturbance input are assumed to be of one of two types namely: (i) they are assumed to satisfy a homogenous differential equation, for example sinusoidal signals, or (ii) they are assumed to satisfy some given norm bound, for
example on energy. Both linear and nonlinear systems are considered. However, the results for the nonlinear case are of a preliminary nature.
The book provides a useful introduction to these questions which lie at the core of control. A brief introduction to some of the mathematical tools in current use on these problems is also given. The book would, therefore, be a very useful starting point for anybody wishing to carry out research on this
central issue in control theory.

Contents:
1. Two-person zero-sum games
2. Two-person non-zero-sum games
3. N-person games

This volume is a valuable addition to the renowned New Mathematical Library. It contains a good combination of the pure mathematical theory and the ways in which it can be applied to real problems. Written with students and for them, this is a highly recommendable text.

Title	200% OF NOTHING. An Eye-Opening Tour Through the Twists and Turns of Math Abuse and Innumeracy.
Author	A.K. Dewdney.
Publisher	New York: Wiley, 1993, pp. 182, US$19.95.

Reviewer:
Institute	University of Bath
Place	Bath, U.K.
Name	C. Chatfield

Title	MEDICAL STATISTICS. A Commonsense Approach, 2nd edition.
Author	M.J. Campbell and D. Machin.
Publisher	Chichester, U.K.: Wiley, 1993, pp. ix + 189, US$28.95/,14.95.

Reviewer:
Institute	Fred Hutchinson Cancer Research Center
Place	Seattle, U.S.A.
Name	P. Feigl

Title	DRUG SAFETY ASSESSMENT IN CLINICAL TRIALS.
Author	G. Sogliero-Gilbert (Ed.).
Publisher	New York: Dekker, 1993, pp. x + 437, US$135.00.

Reviewer:
Institute	Harvard University
Place	Cambridge, U.S.A.
Name	S.W. Lagakos

Reviewer:
Institute	University of Sheffield
Place	Sheffield, U.K.
Name	R.M. Loynes

Reviewer:
Institute	University of Oxford
Place	Oxford, U.K.
Name	B.D. Ripley

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

Reviewer:
Institute	Queen's University
Place	Kingston, Canada
Name	J.T. Smith

Reviewer:
Institute	Queen Mary and Westfield College
Place	London, U.K.
Name	D.J. Balding

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

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

Reviewer:
Institute	University of Cape Town
Place	Cape Town, South Africa
Name	J.M. Juritz

Reviewer:
Institute	The Open University
Place	Milton Keynes, U.K.
Name	D.J. Hand

Title	ELEMENTARY PROBABILITY.
Author	D. Stirzaker.
Publisher	Cambridge University Press, 1994, pp. x + 406, ,45.00/US$64.95 Cloth; ,15.95/US$24.94 Paper.

Reviewer:
Institute	University of Southampton
Place	Southampton, U.K.
Name	S.M. Lewis and P. Prescott

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

Reviewer:
Institute	Colorado State University
Place	Ft. Collins, U.S.A.
Name	R.W. Butler

Reviewer:
Institute	University of Cambridge
Place	Cambridge, U.K.
Name	G.A. Young

Reviewer:
Institute	CSIRO
Place	Melbourne, Australia
Name	A.J. Miller

Reviewer:
Institute	University of Leeds
Place	Leeds, U.K.
Name	C.C. Taylor

Reviewer:
Institute	University of California
Place	Berkeley, U.S.A.
Name	E.L. Lehmann

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

Reviewer:
Institute	University of North Carolina
Place	Chapel Hill, U.S.A.
Name	R.L. Smith

Reviewer:
Institute	University of Manchester Institute of Science and Technology
Place	Manchester, U.K.
Name	T. Subba Rao

Reviewer:
Institute	Imperial College of Science, Technology and Medicine
Place	London, U.K.
Name	C.J. Ridler-Rowe

Reviewer:
Institute	Limburgs Universitaire Centrum
Place	Diepenbeek, Belgium
Name	N. Veraverbeke

Reviewer:
Institute	University of Washington
Place	Seattle, U.S.A.
Name	R. Pyke

Reviewer:
Institute	University of Exeter
Place	Exeter, U.K.
Name	D.K. Smith

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

Reviewer:
Institute	University of St. Andrews
Place	St. Andrews, U.K.
Name	A.W. Kemp

Reviewer:
Institute	Madison, Wisconsin
Place	U.S.A.
Name	C.A. Fung

Reviewer:
Institute	University of East Anglia
Place	Norwich, U.K.
Name	T. Lewis

Reviewer:
Institute	University of Essex
Place	Colchester, U.K.
Name	G.A. Barnard

Reviewer:
Institute	Oklahoma State University
Place	Stillwater, U.S.A.
Name	J.L. Folks