Short Book Reviews On-Line 1992

SCIENCE, IDEOLOGY, AND THE MEDIA. The Cyril Burt Scandal.	R. Fletcher.
MEDICAL USES OF STATISTICS, 2nd edition.	J.C. Bailar III and F. Mosteller (Eds.).
MACSYMA FOR STATISTICIANS.	B. Heller.
FUNDAMENTALS OF EXPLORATORY ANALYSIS OF VARIANCE.	D.C. Hoaglin, F. Mosteller and J.W. Tukey (Eds.).
STATISTICS IN RESEARCH AND DEVELOPMENT, 2nd edition.	R. Caulcutt.
INTRODUCTION TO RELIABILITY ANALYSIS: PROBABILITY MODELS AND STATISTICAL METHODS.	S. Zacks.
INFERENZA STATISTICA.	A. Azzalini.
DESIGN AND INFERENCE IN FINITE POPULATION SAMPLING.	A.S. Hedayat and B.K. Sinha.
SAMPLE SURVEY PRINCIPLES AND METHODS.	V. Barnett.
MEASUREMENT ERRORS IN SURVEYS. Edited by P. Biemer,	R. Groves, L. Lyberg, N. Mathiowetz and S. Sudman.
SURVEY SAMPLING PRINCIPLES.	E.K. Foreman.
MODEL ASSISTED SURVEY SAMPLING.	C.E. Sarndal, B. Swensson and J. Wretman.
ANALYSIS OF VARIANCE IN EXPERIMENTAL DESIGN.	H.R. Lindman.
A USER'S GUIDE TO PRINCIPAL COMPONENTS.	J.E. Jackson.
APPLIED MULTIVARIATE DATA ANALYSIS.	B.S. Everitt and G. Dunn.
MORPHOMETRIC TOOLS FOR LANDMARK DATA. Geometry and Biology.	F.L. Bookstein.
META-ANALYSIS BY THE CONFIDENCE PROFILE METHOD.	D.M. Eddy, V. Hasselblad and R.Schachter.
THE BOOTSTRAP AND EDGEWORTH EXPANSION.	P. Hall.
STOCHASTIC CURVE ESTIMATION.	M. Rosenblatt.
CONTINUOUS MARTINGALES AND BROWNIAN MOTION.	D. Revuz and M. Yor.
UNCERTAINTY AND VAGUENESS IN KNOWLEDGE BASED SYSTEMS. NUMERICAL METHODS.	R. Kruse, E. Schwecke and J. Heinsohn.
EVOLUTION OF RANDOM SEARCH TREES.	H.M. Mahmoud.
ALGEBRAIC METHODS FOR SIGNAL PROCESSING AND COMMUNICATIONS CODING.	R.E. Blahut.
COMMUNICATION THEORY.	C.M. Goldie and R.G.E. Pinch.
SPECTRA OF RANDOM AND ALMOST-PERIODIC OPERATORS.	L. Pastur and A. Figotin.
INTEGER PROGRAMMING.	S. Walukiewicz.
OPTIMAL CONTROL THEORY AND STATIC OPTIMIZATION IN ECONOMICS.	D. Leonard and N. Van Long.
SLIDING MODELS IN CONTROL OPTIMIZATION.	V.I. Utkin.
GLIMPSES OF INDIA'S STATISTICAL HERITAGE.	J.K. Ghosh, S.K. Mitra and K.R. Parthasarathay (Eds.).
PROBABILITY AND THE ART OF JUDGEMENT.	R. Jeffrey.
A PRACTICAL GUIDE TO STATISTICAL QUALITY IMPROVEMENT: OPENING UP THE STATISTICAL TOOLBOX. M.R. Beauregard,	R.J. Mikulak and B. Olson.
STATISTICAL QUALITY DESIGN AND CONTROL: CONTEMPORARY CONCEPTS AND METHODS.R.E.	DeVor, T.-H. Chang and J.W. Sutherland.
BIVARIATE DISCRETE DISTRIBUTIONS.	S. Kocherlakota and K. Kocherlakota.
DISCRETE STOCHASTICS.	K. Jacobs.
FUNDAMENTALS OF BIOSTATISTICAL INFERENCE.	C.T. Le.
THE ANALYSIS OF CONTINGENCY TABLES, 2nd edition.	B.S. Everitt.
COMPUTER-INTERACTIVE DATA ANALYSIS.	A.D. Lunn and D.R. McNeil
THE ANALYSIS OF STOCHASTIC PROCESSES USING GLIM.	J.K. Lindsey.
REGRESSION ANALYSIS BY EXAMPLE, 2nd edition.	S. Chatterjee and B. Price
REGRESSION WITH GRAPHICS. A Second Course in Applied Statistics.	L.C. Hamilton.
HIERARCHICAL LINEAR MODELS: APPLICATIONS AND DATA ANALYSIS METHODS.	A.S. Bryk and S.W. Raudenbush.
CORRESPONDENCE ANALYSIS HANDBOOK.	J.-P. Benzecri.
DISCRIMINANT ANALYSIS AND STATISTICAL PATTERN RECOGNITION	G.J. McLachlan.
STATISTICS FOR SPATIAL DATA.	N.A.C. Cressie.
CONFIDENCE INTERVALS ON VARIANCE COMPONENTS.	R.D. Burdick and G.A. Graybill.
VARIANCE COMPONENTS.	S.R. Searle, G. Casella and C.E. McCulloch.
DESIGNS AND THEIR CODES.	E.F. Assmus Jr. and J.D. Key.
QUADRATIC FORMS IN RANDOM VARIABLES. Theory and Applications.	A.M. Mathai and S.B. Provost.
EXPLORING THE LIMITS OF BOOTSTRAP.	R. LePage and L. Billard (Eds.).
ARMA MODEL IDENTIFICATION.	B.S. Choi.
FOUNDATIONS OF THE PREDICTION PROCESS.	F.B. Knight.
THE ECONOMETRIC ANALYSIS OF TRANSITION DATA.	T. Lancaster.
TEN LECTURES ON WAVELETS.	I. Daubechies
RANDOM NUMBER GENERATION AND QUASI-MONTE CARLO METHODS.	H. Niederreiter.
UNIVERSAL ALGEBRA FOR COMPUTER SCIENTISTS.	W. Wechler.
RANDOM SERIES AND STOCHASTIC INTEGRALS: SINGLE AND MULTIPLE.	S. Kwapieñ and W. Woyczyñski.
IDENTIFIABILITY IN STOCHASTIC MODELS. CHARACTERIZATION OF PROBABILITY MODELS	B.L.S. Prakasa Rao.
STOCHASTIC APPROXIMATION AND OPTIMIZATION OF RANDOM SYSTEMS.	L. Ljung, G. Pflug and H. Walk.
STOCHASTIC CONTROL OF PARTIALLY OBSERVABLE SYSTEMS.	A Bensoussan.

Contents:
1. Case for the Defence
PART I : Preamble; Introduction; Approach; Jury; Prosecution; Cross-Examination; Fraud? True or False?; Rehabilitation; Relevance?
PART II : Witnesses for the Prosecution. Cross- Examination Introduction
2. The BBC: The Intelligence Man; The billing; Misrepresentations; ....; Questions requiring public answers; Notes
3. Professor Leon Kamin; General Charges; The 1943 paper "Ability and Income"; The "Invariant Correlations"; ....; Notes
4. Dr. Oliver Gillie and the Press; The Most Sensational Charge of Fraud This Century; ...; The Missing Ladies - and Parapsychology; ...; Notes
5. The Clarkes and Hull; Collaboration in the Public Charge of Fraud; The Famous Articles:...: Notes
6. Professor Leslie Hearnshaw: The Official Biography;... Dorfman and Stigler on Burt's Statistics; ...; Strange Conjectures; ...; Notes
7. Professors Jensen and Eysenck; Questions Requiring Public Answers; Notes
8. The Sociologists of Education; ...; Heredity and Environment: Burt; Heredity and Environment: The Sociologists of Education; Conclusion; Notes
9. Final summary note
PART III: Malfeasance and Fraud? Summing Up, Additional Evidence, and Verdict
10. Malfeasance?; Eugenics; Racism; Social Class; ...; Note
11. Fraud? The Minor Charges; ...; Falsifying the History of Factor Analysis
12. Fraud? The Major Charges; ...; The Vulnerability of Exactitude; Age; Notes
13. Verdict
PART IV : Questions and Conclusions
14. Remaining considerations:...; Notes

This volume gives a detailed rebuttal of the charges made in 1976 and later to the effect that the British psychologist Sir Cyril Burt faked his data and committed other scientific crimes and misdemeanors. Following the publication of this book, and that by Joynson reviewed in December 1989, a group of members of the British Psychological Society (BPS) asked that the Society should invite the Royal Society to institute a formal inquiry in order to reconsider Burt's conduct. The BPS issued a unanimously agreed statement on the 24th February 1992 summarized as follows:
1. Twelve years ago the Society assumed that Burt was guilty of fraud. We no longer hold that view. It is not that we have looked at the evidence and come to a different conclusion. We have decided that we ought not to be taking a corporate position at all.
2. The Society has procedures for investigating professional misconduct by members. They do not, and cannot, apply to deceased members.
3. The allegations concerning Burt are a matter of historical debate. It is not the job of a learned body in a free society to adjudicate on such matters. Our role is to promote discussion so that people can make up their own minds.
4. To this end, the Council will organize a major sym-posium, chaired by the President, at its next London Conference, in December this year, at which the alle-gations directed at Burt will be reviewed.
The Council resolution condemning Burt can be found on p.71 of the Monthly Bulletin of the BPS for February 1980.
In the present book one collector of Burt's data is referred to as Miss Lettice Ramsey. She was, alas, for many years widow of Frank Ramsey, an origin-ator of the modern theory of personal probability.

Contents:
PART I : Broad Concepts and Analytic Techniques
1. Statistical concepts fundamental to investigations
2. Some uses of statistical thinking
3. Use of statistical analysis in the New England Journal of Medicine
PART II : Design
4. Designs for experiments: Parallel comparisons of treatment
5. Crossover and self-controlled designs in clinical research
6. Studies without internal controls
7. The series of consecutive cases as a device for assessing outcomes of interventions
8. A classification for biomedical research reports
PART III: Analysis
9. Decision analysis
10. P values
11. Simple linear regression in medical research
12. Comparing the means of several groups
13. Analyzing data from ordered categories
14. Statistical analysis of survival data
15. Contingency tables in medical studies
PART IV : Communicating Results
16. Guidelines for statistical reporting in articles of medical journals
17. Reporting on methods in clinical trials
18. Statistical consultation in clinical research: A two-way street
19. The importance of beta, the Type II error, and the sample size in the design and interpretation of the randomized controlled trial: Survey of two sets of "negative" trials
20. Writing about numbers
PART V : Reviews and Meta-Studies
21. Medical technology assessment
22. Combining results from independent investigations: Meta-analysis in clinical research
23. Meta-analyses of randomized control trials: An update of the quality and methodology

In the late 1970s, the editors of the New England Journal of Medicine encouraged a group of Harvard-based statisticians to survey the current use of statistics in their journal and to recommend future practice. The first edition of this book (1986) [Short Book Reviews, Vol. 7, p.3] contained, in the main, articles written by the group and published in the journal. The second edition updates the surveys and the recommendations: seven chapters are new, four dropped, two completely rewritten, and the rest revised. Many of the chapters are models of clear writing; the chapter on P values is a masterly treatment of an elementary but complex topic. The book gives sound, practical advice and helps the reader to understand the purposes of statistical techniques, but is not a text book and does not explain the details of the techniques. I know of no other book which describes "what every statisticians knows" so clearly and succinctly.

Contents:
1. Getting started
2. Variables, lists, equations, functions, and arrays
3. Iteration, conditionals, blocks, and recursion
4. Part selection, substitution, and evaluate
5. Internal representation, storage, general utilities
6. Matrices and lists
7. Advanced uses of MACSYMA

Readership: Professional statisticians, probabilists, mathematicians and students

This book provides an introduction to the basic principles and ideas of MACSYMA, a computer sys-tem for symbolic and numerical calculation. The book has numerous exercises and may be used as a text.
MACSYMA was one of the first systems for symbolic computation to receive widespread use. Since its development, other systems including Mathematica, Maple and Reduce have been introduced. Although these systems all differ in their structure and use there is much that is common. This book is a useful introduction to the field. Although much of the book is devoted to general techniques such as for handling matrices and lists, there are examples which indicate the power of symbolic calculation. The section on Edgeworth series shows how the series may be derived from first principles and used in particular cases. The section on partitions leads to the calculations of cumulants from moments and in an exercise, k-statistics. These are interesting ideas and challenging calculations.
The book is an introduction, designed to be accessible by students. It is particularly useful for those using MACSYMA. Research statisticians will wish for less linear algebra and more development peculiar to statistics. There is a need for this and for more advanced books.

Contents:
1. Concepts and examples of analysis of variance
2. Purposes of analyzing data that come in a form inviting us to apply tools from the analysis of variance
3. Preliminary examination of data
4. Types of factors and their structural layouts
5. Value-splitting: taking the data apart
6. Value splitting involving more factors
7. Mean squares, F tests, and estimates of variance
8. Graphical display as an aid to analysis
9. Components of variance
10. Which denominator?
11. Assessing changes
12. Qualitative and quantitative confidence
13. Introduction to transformation

This third volume by the same editors, with contributions by eleven authors, is something of a surprise: unlike its predecessors in the series, it is centered round a classical, confirmatory technique. However, while covering all the traditional topics in mainly balanced analysis of variance, the authors write from a perspective that is distinctly exploratory data analysis, and to excellent advantage.
The result is a self-contained introduction to standard ANOVA, accessible to a wide range of backgrounds, and offering a fresh understanding of the subject enriched by data-motivation of the models and methods. The unavoidable message, crucial in an age of computational sophistication in the hands of everyone, is that the analysis begins and ends with the data in its problem context and the standard computations enshrined in computer packages are only some of the tools required for solution of the problem. Some new (or newly exposed) graphical techniques will be appreciated by even experienced users of ANOVA. To mention just two, side-by-side plots of effects and residuals provide a powerful visual counterpart to the ANOVA table, and interference plots are a graphical accompaniment to the studentized range approach to multiple comparisons.

Contents:
1. What is statistics?
2. Describing the sample
3. Describing the population
4. Testing and estimation: One sample
5. Testing and estimation: Two samples
6. Testing and estimation: Qualitative data
7. Testing and estimation: Assumptions
8. Statistical process control
9. Detecting process changes
10. Investigating the process - an experiment
11. Why was the experiment not successful?
12. Some simple but effective experiments
13. Adapting the simple experiments
14. Improving a bad experiment
15. Analysis of variance
16. An introduction to Taguchi techniques

Readership: Anyone concerned with the study of complex processes in any industry. Some facility with algebra is needed, but no calculus is required, of the reader

This second edition of the 1983 book by the same title includes new sections on statistical process control and Taguchi methods. The book is friendly to read. Its selection of topics is standard, but I enjoyed it more than the usual introductory text because of its use, in the early chapters, of a single manufacturing example to set a common context for all the methods presented.
The first nine chapters deal with finding out whether something has changed, with decisions supported by significance tests, thus, the traditional role of statistics as passive arbiter. The remaining chapters introduce designed experimentation to show the strength of statistics as aggressive investigator, the analysis of variance, and Taguchi's contributions. Although the early chapters are useful "for later", I might advise readers who are driven by R&D to read Chapters 10, 11, and 12 first, to find out why it's worthwhile to learn the rest of statistics. To quote from the book's pre-face, "You don't know what you are missing until you have read it all."
The book is written for practitioners in industry, but university students who plan a career in industry, or in consulting, will also find it useful.

Contents:
1. System effectiveness
2. Life distributions, models and their characteristics
3. Reliability of composite systems
4. Reliability of repairable systems
5. Graphical analysis of life data
6. Estimation of life distributions and system characteristics
7. Maximum likelihood estimators and confidence intervals for specific life distributions
8. Bayesian reliability estimation and prediction
9. Reliability demonstration: Testing and acceptance procedures

The present monograph grew out of a workshop on statistical methods of reliability analysis for engineers. It aims at bringing the engineer, student or practitioner, up to date on modern reliability methodology from the point of view of probabilistic modelling and statistical estimation. The author explains very clearly the practical background to reliability questions as well as the theory needed in solving these questions. Most of the theoretical results are presented without proofs. This choice is justified for the intended audience, as the author puts it: "the book concentrates on the methodology of the subject and on understanding theoretical results rather than on its theoretical development." On the other hand, a vast number of examples is worked out in full detail. These, together with numerous interesting exercises form the core of this excellent text. I am convinced that many engineering students will find this a very readable introduction to reliability.

Contents:
1. Introduction
2. Elements of probability theory
3. Likelihood
4. Maximum likelihood estimates
5. Hypothesis testing
6. Linear models

This book reflects the contents of a course given by the author at the University of Padua to undergraduate students in the Statistical Sciences Department. Its aim is to present the main concepts of classical statistical inference in an unifying frame-work via the likelihood principle. The intuitive arguments underlying the methods derived from the likelihood principle are clearly stated and yet the mathematical content of the book is rigorous. Indeed Chapter 2 in itself is an extremely useful compilation of basic elements of probability theory.
The author succeeds in his aim of maintaining a lean and interesting thread through the book. The layout of the book is also excellent as keywords mark the margins of the text. Every chapter contains a set of exercises. Most of these are quite complex, possibly a hint to the standard required to pass the exam.

Contents:
1. A unified setup for probability sampling
2. Inference in finite population sampling
3. The Horvitz-Thompson estimator
4. Simple random and allied sampling designs
5. Use of auxiliary size measures in survey sampling: Strategies based on probability proportional to size schemes of sampling
6. Uses of auxiliary size measures in survey sampling: Ratio and regression methods of estimation
7. Cluster sampling designs
8. Systematic sampling designs
9. Stratified sampling designs
10. Superpopulation approach to inference in finite population sampling
11. Randomized response: A data-gathering tool for sensitive characteristics
12. Special topics: Small area estimation, nonresponse problems, and resampling techniques

This is one of a new generation of books on survey sampling. Despite its title it is not about inference. Issues of recent concern, such as the central limit theorem, consistency, conditional inference, description and analysis, are not mentioned. The book is about the choice of sampling strategies, design and estimator pairs, for univariate surveys. Many fine points of detail are carefully considered. The failure to consider designs, other than randomized response, for estimating measurement errors was particularly disappointing. Do not throw your copy of W.G. Cochran's Sampling
Techniques away just yet.

Contents:
1. Introduction
2. Simple random sampling
3. Carrying out a sample survey
4. Ratios: Ratio and regression estimators
5. Stratified populations and stratified simple random sampling
6. Cluster and multi-stage sampling
7. Postscript

Readership: Advanced undergraduate and beginning postgraduate students in statistics. Practitioners in any subject requiring a good knowledge of survey methods

This is a revised and expanded version of Elements of Sampling Theory, which has been consistently popular since it was first published in 1974. There is more material on such basic methodological topics as ratio estimation and multi-stage sampling and a good discussion of non-sampling errors and the problems of carrying out a survey in practice. There is also a welcome change to a more standard notation. The new version should prove even more valuable than the original.

Contents:
Introduction
1. Measurement error across disciplines
PART I : The Questionnaire
2. The current status of questionnaire design
3. Response alternatives: The impact of their choice and presentation order
4. Context effects in the general social survey
5. Mode effects of cognitively designed recall questions: A comparison of answers to telephone and mail surveys
6. Nonexperimental research on question wording effects: A contribution to solving the generalizability problem
7. Measurement errors in business surveys
PART II : Respondents and Responses
8. Recall error: Sources and bias reduction techniques
9. Measurement effects in self vs. proxy response to survey questions: An information-processing perspective
10. An alternative approach to obtaining personal history data
11. The item count technique as a method of indirect questioning: A review of its development and a case study application
12. Toward a response model in establishment surveys
PART III: Interviewers and Other Means of Data Collection
13. Data collection methods and measurement error: An overview
14. Reducing interviewer-related error through interviewer training, supervision, and other means
15. The design and analysis of re-interview: An overview
16. Expenditure diary surveys and their associated errors
17. A review of errors of direct observation in crop yield surveys
18. Measurement error in continuing surveys of the grocery retail trade using electronic data collection methods
PART IV : Measurement Errors in the Interview Process
19. Conversation with a purpose - or conversation? Interaction in the standardized interview
20. Cognitive laboratory methods: A taxonomy
21. Studying respondent-interviewer interaction: The relationship between interviewing style, interviewer behavior, and response behavior
22. The effect of interviewer and respondent characteristics on the quality of survey data: A multilevel model
23. Interviewer, respondent, and regional office effects on response variance: A statistical decomposition
PART V : Modeling Measurement Errors and their Effects on Estimation and Data Analysis
24. Approaches to the modeling of measurement errors
25. A mixed model for analyzing measurement errors for dichotomous variables
26. Models for memory effects in count data
27. Simple response variance: Estimation and determinants
28. Evaluation of measurement instruments using a structural modeling approach
29. A path analysis of cross-national data taking measurement errors into account
30. Regression estimation in the presence of measurement error
31. Chi-Squared tests with complex survey data subject to misclassification error
32. The effect of measurement error on event history analysis

This book will be an aid to survey statisticians and to research workers who must work with survey data. It is a collection of papers providing a thorough account of the current state of research on measurement errors in surveys and reporting on the findings of new research, as well as on interdisciplinary approaches in modeling, assessing and reducing measurement errors. It examines current issues in questionnaire design, the respondent-interviewer relationship, modeling measurement errors, and their effects on estimation and data analysis.

Contents:
1. Introduction
2. Some basic concepts
3. Simple random sampling and unbiased estimates
4. Ratio and regression estimates and estimates of ratios
5. Stratified sampling
6. Selection with probability proportional to size
7. Multistage sampling: Concept
8. Multistage sampling: Application
9. Further sampling techniques
10. Control of nonsample errors
11. Sample design
12. Sample frames, selection systems, and master samples
13. Household survey master samples
14. Survey design
15. Design of series of surveys

Readership: Survey statisticians, graduate students in statistics and social sciences

This book is a welcome addition to the survey sampling library. No, or very little, competence in statistical theory on the part of the reader is assumed; proofs are in general omitted and the emphasis is on survey applications. The numerical examples are based on a very small artificial population, "Survey-town", for which the arithmetic can be easily verified. Chapters 1 to 8 provide a fairly standard coverage for a survey sampling course; Chapter 9 deals with more specialist topics; and Chapters 10 through 15 are much more a text in general survey design and execution is-sues. The references and reading for each chapter help-fully give the relevant page numbers when books are cited.
Throughout the book Foreman's considerable knowledge and experience give the material a flavour of authenticity that makes the material digestible, despite some daunting, for non-statisticians, algebra in the middle chapters. This is a book in the tradition of Hansen, and of Kish, who, with others, have streets named after them in "Surveytown".

Contents:
PART I : Principles of Estimation for Finite Populations and Importance Sampling Designs
1. Survey sampling in theory and practice
2. Basic ideas in estimation from probability samples
3. Unbiased estimation for element sampling designs
4. Unbiased estimation for cluster sampling and sampling in two or more stages
5. Introduction to more complex problems
PART II : Estimation Through Linear Modelling, Using Auxiliary Variables
6. The regression estimator
7. Regression estimators for element sampling designs
8. Regression estimators for cluster sampling and two- stage sampling
PART III: Further Questions in Design and Analysis of Surveys
9. Two-phase sampling
10. Estimation for domains
11. Variance estimation
12. Searching for optimal sampling designs
13. Further statistical techniques for survey data
PART IV : A Broader View of Errors in Surveys
14. Nonsampling errors and extensions of probability sampling theory
15. Nonresponse
16. Measurement errors
17. Quality declarations for survey data

Readership: Survey statisticians, graduate students in statistics and social sciences

The title of this book in itself represents a bid to reclaim the word "model" in survey sampling from those who advocate model-based inference. This relatively advanced text is based on the design-based (randomization theory) approach; it does, however, take on board the considerable influence of statistical modelling on survey sampling in recent years. One of its strengths is in the way it draws on statistical modelling in its derivation of estimators. The text shows an appreciation of practical issues, though these are not generally elaborated in non-technical terms. The overall perspective, based on unequal probability sampling, is useful in integrating the treatment. The authors manage to include a coherent treatment of measurement error and nonresponse within their overall framework. This book is a very useful addition to the survey sampling literature, one that will be most useful for those with a strong statistical grounding.

Contents:
1. Review of statistical concepts
2. Analysis of variance, one-way, fixed effects
3. Comparing groups
4. Other multi-comparison methods
5. Two-way analysis of variance
6. Random effects
7. Higher-way designs
8. Nested designs
9. Other incomplete designs
10. One-way designs with quantitative factors
11. Trend analyses in multi-factor designs
12. Basic matrix algebra
13. Multivariate analysis of variance
14. Analysis of covariance
15. General linear model

This is a book to dip into; it is not for reading from cover to cover. It attempts to explain to the non-statistician how to analyze a number of commonly occurring experimental designs and provides the reader with a wealth of numerical examples. Attention is strongly focused on the normal theory linear model. The author examines the assumptions underlying this model, although there is little guidance on what to do if these assumptions are violated. After a two-page introduction to probability and statistics, we are presented with a section on skewness and kurtosis which would leave the less mathematically trained thinking that there might not be anything in the book for them.
The book is well laid out and easy to follow, but may lure the experimenter into thinking that problems with data can be pigeon-holed into, for example, one-way analysis of variance, two-way analysis of variance, etc. A chapter giving some preliminary advice on looking at data and the consideration of suitable models would have been an advantage.

Contents:
1. Getting started
2. PCA with more than two variables
3. Scaling of data
4. Inferential procedures
5. Putting it all together - Hearing loss I
6. Operations with group data
7. Vector Interpretation I: Simplifications and inferential techniques
8. Vector Interpretation II: Rotation
9. A Case History - Hearing loss II
10. Singular value decomposition: Multidimensional scaling I
11. Distance models: Multidimensional scaling II
12. Linear models I: Regression, PCA of predictor variables
13. Linear models II: Analysis of variance; PCA of response variables
14. Other applications of PCA
15. Flatland: Special procedures for two dimensions
16. Odds and ends
17. What is factor analysis anyhow?
18. Other competitors

Readership: Statisticians, advanced students of statistics, experimental scientists

Principal component analysis has many different uses, as is readily seen from this book. The bibliography has fifty-four pages, containing many references which have appeared since I.T. Jolliffe's Principal Component Analysis [Short Book Reviews, Vol. 6, p.45], for example in areas such as influence, partial least squares and multiple correspondence analysis. The book is designed for users; there are very few theorems, and only one proof. Techniques are illustrated on twenty-five data sets, several resulting from the author's experience as an employee with Kodak. Clear practical advice frequently flows from this experience. Particularly attractive is the introduction of new procedures through simple worked examples. There are no exercises, or guides to computer package facilities. The book ends with "... although it may appear that the pioneers in the field have skimmed off all the cream, there are still plenty of development opportunities to make PCA an even more useful technique. The increase in articles during the past few years bears this out. Come aboard." This is typical of the book's lucid style which will result in many readers reaching the end, and this conclusion.

Contents:
PART I : Approaches to Analysing Data
1. Data and statistics
2. Mathematical and statistical background
PART II : Exploring Multivariate Data
3. The initial examination of multivariate data
4. Reducing the dimensionality of multivariate data: Principal components and correspondence analysis
5. Multidimensional scaling
6. Cluster analysis
PART III: Regression Models
7. The generalised linear model
8. Regression and the analysis of variance
9. Linear models for categorical data
10. Models for rates and survival times
11. Analysis of repeated measures
12. Discriminant analysis
PART IV : Latent Variable Models
13. Factor analysis
14. Covariance structure models
Readership: Students and research workers in the social and behavioural sciences

This book is an extensive revision of Advanced Methods of Data Exploration and Modelling, by the same authors. It is assumed that readers have "a firm grasp of basic statistics and mathematics". However, the authors deliberately avoid mathematical development; for example standard principal component results are stated without proof. Although wide-ranging, the book is easy to read. An attractive feature is the large number of real examples, many drawn from the authors' practical experience. References are up-to-date, and new developments are critically discussed. A minor cavil is that here, as is often the case, single-link cluster analysis is presented by a more complex algorithm than is necessary. Each chapter, except for Chap-ter 1, has exercises, some with solutions. The Appendix on computing covers just one and one-half pages. This is a very useful book which can be thoroughly recommended.

Contents:
1. Introduction
2. Preliminaries
3. Landmarks
4. Distance measures
5. Shape coodinates
6. Principal axes of shape change for triangles
7. Features of shape comparison
8. Retrospect and prospect

Morphometrics is the statistical study of biological shape and shape change. Its richest data are landmarks which are points such as the "tip of nose" that have biological names as well as geometric locations. This book is the first symstematic survey of morphometric methods for landmark data. It has an excellent layout incorporating a number of extremely helpful diagrams. Professor Bookstein has himself pioneered various techniques and concepts, and has brought forward the ideas of D'Arcy Thompson into a modern formulation. The book will certainly be a landmark volume in this difficult but important field of shape analysis.

Contents:
1. Examples of problems
2. Setting up the analysis
3. Formulas of the confidence profile method
4. Solutions to the example problems
5. Implementation issues
6. Conclusions

Among many features of the book are the following: (i) advocacy of the confidence profile method, which consists in writing down a full probability model for the system under study and thence computing either maximum likelihood estimates and the corresponding matrix of observed second derivatives or, when a prior distribution is available, computing a posterior distribution for the parameter of interest;
(ii) use of influence diagrams to represent the structure of probability models; (iii) a careful discussion of the types of error, systematic and random, likely to be encountered when evidence from a number of different sources is assembled bearing on a particular medical issue (although publication bias seems not to be mentioned); (iv) advocacy and illustration of (i) on a considerable variety of interesting examples; (v) the assertion that "the main limitation of classical (statistical) methods is that they address individual experiments or discrete sets of data and do not adjust for biases" (p.351) and the suggestion that "the notion of adjusting for biases and...are new to the Confidence Profile Method".
Opinions on many of these points will vary. Your reviewer was unable to see anything new in (i), found the influence diagrams confusing rather than helpful (although more experience might well lead to a change of opinion), enjoyed the qualitative discussion of the examples, was disappointed at the quantitative discussion which often seemed to rely on untested numerical assumptions, and, while recognizing that overviews are very important and difficult to achieve satisfactorily, was unconvinced that any new statistical principles are involved.

Contents:
1. Principles of bootstrap methodology
2. Principles of Edgeworth expansion
3. An Edgeworth view of the bootstrap
4. Bootstrap curve estimation
5. Details of mathematical rigour

This is an authoritative book-length discussion of bootstrap theory by one of its main architects; an important work that is likely to be widely read and referenced. The author states that the aim of the book is to address two quite different topics, the bootstrap on the one hand and the Edgeworth expansion on the other, in the belief that each can shed light on the other. He hits this target admirably, and though some of the material is technical, its treatment is clear, highly informative, and scholarly but unfussy. Although methods of simulation are discussed in an appendix, this is not primarily a book about applying the bootstrap, but it provides theoretical underpinning for a wide range of applications. This is a book which every university statistics library should buy, and which researchers and regular users of the bootstrap will want to own and to consult regularly.

Contents:
1. Origins
2. Local asymptotics
3. Global measures of deviation
4. Cross-validation
5. Measures of short-range dependence
6. Probability density and regression estimation in the case of short-range dependence
7. Spectral densities and cumulants
8. Examples of long-range dependence
9. Curve estimation and long-range dependence
10. Open questions

This monograph contains the notes of a series of ten lectures given at the University of California, Davis, during June 1989. The theme of the lectures is: curve estimation, based on independent and dependent observations. The book starts with some historical comments on the early examples in this area: estimation of the spectral density of a stationary sequence. Remarkably enough these first examples were related to dependent observations, but the main stream of results afterwards were for models with independent observations. Chapters 2-4, which cover about half of the book, deal with asymptotic theory under independence for density estimation and regression function estimation, based on kernels or on splines. Also the important question of bandwidth choice is discussed briefly. In the remaining chapters, it is discussed to what extent the asymptotic results still hold under appropriate conditions of short-range and long-range dependence of the sequence of observations. The bibliography contains only about ninety items. The author did of course not take up the impossible task of covering the whole literature on the subject, but he suceeded in sketching the most typical results in this important theory. Unfortunately, the book has no subject index.

Contents:
0. Preliminaries
1. Introduction
2. Martingales
3. Markov processes
4. Stochastic integration
5. Representation of martingales
6. Local times
7. Generators and time reversal
8. Girsanov's theorem and first applications
9. Stochastic differential equations
10. Additive functionals of Brownian motion
11. Bessel processes and Ray-Knight theorems
12. Limit theorems in distribution

Brownian motion is a special case of many classes of stochastic process. It is a continuous martingale, a Gaussian process, and a Markov process (indeed it is a diffusion and a process with independent increments). The philosophy of the authors is that its study provides the ideal setting in which to exhibit the range of techniques of modern ("continuous") probability, both in the abstract and as valuable tools for many calculations. In spite of the book's title, each of the chapters treats its subject at a general level. Brownian motion recurs as an illustrative special case, and motivation, rather than as a limit to ambition. The treatment is extremely wide ranging, thorough and rigorous. While there are few formal pre-requisites beyond a background in measure theoretic probability, the book requires considerable mathematical sophistication, and some of the material, though well motivated, is necessarily quite technical. There
are lots of exercises, and the well-equipped reader should find it enjoyable, rewarding and enlightening.

Contents:
1. General considerations of uncertainty and vagueness
2. Introduction
3. Vague data
4. Probability theory
5. Random sets
6. Mass distributions
7. On graphical representations
8. Modeling aspects
9. Heuristic models
10. Fuzzy set based models
11. Reasoning with L-sets
12. Probability based models
13. Models based on the Dempster-Shafer theory of evidence
14. Reasoning with mass distributions
15. Related research

Readership: Graduate students in artificial intelligence, operations research, and applied probability. Researchers and practitioners in knowledge based systems

The aim of the book is to describe methods for handling ignorance within knowledge based systems. It distinguishes between uncertainty and vagueness, the former referring to a person's confidence in an item of data, and the latter to ignorance about the precision of data. It discusses formal representations of vagueness and uncertainty and compares these with existing approaches, focusing on three: probabilistic reasoning, fuzzy reasoning, and evidential reasoning.
The first six chapters outline the mathematical theories, emphasizing formal properties so as to provide a basis for what follows. The remaining chapters apply the mathematics to modelling uncertainty and vagueness.
The book is restricted to numerical methods based on measure-theoretic concepts; it does not discuss logical approaches except in the final chapter, which outlines nonstandard logics, the integration of uncertainty calculus and logic, and symbolic methods.
It has an extensive reference list.
The length and technical level of the book means that it is not to be undertaken lightly, but it does provide a good grounding in the subject. The early mathematical chapters mean that it is essentially self-contained.

Contents:
1. Basic tools
2. Naturally growing binary search trees
3. Search trees with higher branching factors
4. Trees for multidimensional data
5. Tries
6. Digital search trees

Readership: Professionals and students in computer science and other areas where random tree-like models are sought

The aim of this book is to bring together work on the probabilistic analysis of search trees. The first chapter is simply a collection of the mathematical topics needed later in the book, without any real attempt at unification. The second chapter begins by showing how a binary tree can be constructed from an arbitrary sequence of input symbols, the randomness in the order of this sequence determining the random structure of the tree. This chapter includes discussions of path lengths and extreme paths. Chapters 3 and 4 provide generalizations of binary trees: first to trees with more than two branches at each node and secondly to the particular case of this arising when each node corresponds to a multivariate key and so may be split in multiple ways, yielding a 'quadtree'. Chapter 5 is concerned with 'tries'. A trie stores sets of keys consisting of sequences of symbols and a search for a key is similar to locating a word in a dictionary. Finally, Chapter 6 describes a mixture of search trees and tries.
The book would have benefited from more detailed discussion of the links and relationships between methods, and the contexts in which the different approaches are valuable. It does, however, provide a sound review of work in the area.

Contents:
1. Introduction
2. Mathematical fundamentals
3. Sequences and spectra
4. Cyclic codes and related codes
5. Fast algorithms for convolution
6. Solving Toeplitz systems
7. Fast algorithms for the Fourier transform
8. Decoding of cyclic codes

According to the author the "primary purpose of this monograph is to explore the ties between digital signal processing and error-control codes, with the thought of eventually making them the two components of a unified theory, or of making a large part of the theory of error-control codes a subset of digital signal processing. By studying the properties of the Fourier transform in an arbitrary field, a perspective emerges in which the two subjects are unified." The discrete Fourier transform, over finite and infinite fields, and its applications to coding, to convolution, and to related topics are the central theme in this monograph.

Contents:
0. Introduction
1. Economical representations: Noiseless coding
2. Properties of a message source
3. Reliable transmission
4. Channel coding theorems
5. Error-control codes
6. Cyclic codes
7. Appendix: Rings, fields and vector spaces

Readership: Senior undergraduates and beginning graduate students in mathematics

This is an introductory text in information theory and algebraic coding theory. It deals with noiseless source coding and with coding for the discrete memoryless channel, and for the time-discrete memoryless channel with additive Gaussian noise, in some detail. More attention is given to Markov sources and ergodic sources than might be expected in a book of this length. Topics such as rate-distortion theory, metric entropy, universal coding, and the relation between relative entropy and statistical inference are discussed briefly. Other advanced topics, such as multiple-access channels, are mentioned with appropriate references to the literature. The algebraic coding theory chapters occupy thirty-seven pages. The main topics covered are linear codes, Reed-Muller codes, cyclic codes and BCH codes, and the Hamming and Varsharmov-Gilbert bounds.

Contents:
PART 1: Introduction
PART 2: Asymptotic properties of metrically matrix and differential operators dimensional
PART 3: Integrated density of state in one-problems of second order
PART 4: Asymptotic behavior of the integrated density of states at spectral boundaries in multidimensional problems
PART 5: Lyapunov exponents and the spectrum in one dimension
PART 6: Random operators
PART 7: Almost-periodic operators

There has been over recent years a huge amount of mathematical work devoted to the study of random and almost periodic Schrödinger operators. The present book gives an account of the current knowledge on the sub-ect with a special emphasis in the case of dimension one. The authors have made a special effort to discuss the ideas involved in the proofs and this makes the book pleasant to read. This book will be an aid to people interested by or working in the field.

Contents:
1. Introduction
2. Linear programming
3. Unimodularity and network flows. Cutting-plane methods
4. Branch-and-bound methods
5. The knapsack method
6. Equivalent formulations for integer programs
7. Relaxations of integer problems. Duality
8. Some particular integer programming problems
9. Near-optimal methods
10. Conclusions

This text is principally concerned with mathematical description of methods to solve integer programming problems. The descriptions are clear and well structured. The author is to be congratulated on writing a useful workmanlike book; however, for intense study of the subject and as a research text, no book can approach Integer and Combinatorial Optimization by G. Nemhauser and L. Wolsey, [Short Book Reviews, Vol.9, p. 31].

Contents:
1. Static optimization
2. Ordinary differential equations
3. Introduction to dynamic optimization
4. The maximum principle
5. The calculus of variations and dynamic programming
6. The general constrained control problem
7. Endpoint constraints and transversality conditions
8. Discontinuities in the optimal controls
9. Infinite horizon problems
10. Three special topics

Readership: Mathematical economists, control theorists having an interest in economic applications

This book gives a self-contained introduction to optimal control theory. It is principally aimed at economists but the book would also be of interest to control theorists who wish to see the particular terminology used to describe optimization problems in economics. The book is easy to read yet covers the key facts in the theory of static optimization (Lagrange multipliers, Kuhn-Tucker thoerem, nonlinear programming, etc.) and dynamic optimization (maximum principle, state constraints, end point constraints, Hamilton-Jacobi-Bellman equation etc. The book is recommended to economists who wish to learn (or possibly teach) optimization methods. It is also recommended to control theorists who want to learn some of the language of economists (Shadow prices versus Lag-range multipliers, etc.) In summary, an interesting, readable and clear treatment of optimal control.

Contents:
1. Scope of the theory of sliding modes
2. Mathematical description of motions on discontinuity boundaries
3. The uniqueness problem
4. Stability of sliding modes
5. Singularly perturbed discontinuous systems
6. Decoupling in systems with discontinuous controls
7. Eigenvalue allocation
8. Systems with scalar control
9. Dynamic optimization
10. Control of linear plants in the presence of disturbances
11. Systems with high gains and discontinuous controls
12. Control of distributed parameter plants
13. Control under uncertain conditions
14. State observation and filtering
15. Sliding modes in problems of mathematical programming
16. Manipulator control systems
17. Sliding modes in control of electric motors
18. Examples

Sliding mode control is concerned with the study of differential equations having discontinuous right-hand sides. Often in such systems, the system trajectory is forced onto a surface. The study of the system can then be greatly simplified by introducing an equivalent control action, which causes the trajectory to follow the surface on a sliding mode. This mode of control has found widespread application, especially in systems where the control action is naturally discontinuous, for example, control of electrical machines using power electronic switching devices. The book gives the background theory to sliding-mode control, outlines the effect of model imperfections, presents design methodologies and describes numerous practical examples. Professor Utkin is recognized as the world authority on this topic and hence this English language version of his earlier Russian edition will be warmly welcomed by the control community.

Contents:
1. Remarks on transitive sufficiency, R.R. Bahahdur
2. Learning from counterexamples: At the Indian Statistical Institute in the 1950's, D. Basu
3. My quest for research ideas, V.S. Huzurbazar
4. Random reflections, G. Kallianpur
5. From number theory to national sample survey: An autobiographical letter, D.B. Lahiri
6. Random processes and probabilistic functional analysis: Some contributions, P.R. Masani
7. In statistics by design, K.R. Nair
8. Statistics as a last resort: An autobiographical account, C.R. Rao
9. Combinatorics and I, S.S. Shrikhande
10. Statistics, nutrition, education and social change, P.V. Sukhatme

A substantial number of the world's most influential statisticians in recent decades were born in India, received at least their initial education there and, in many cases, either worked in India or kept substantial contacts there, even if primarily working in other countries. This book celebrates that record by personal statements from ten individuals. The book combines scientific value with considerable human interest.
Your reviewer enjoyed all the essays. It is in some ways invidious to single out particular ones for mention, but the account by the late V.S. Huzurbazar, in particular of his time as H. Jeffreys's only research student in statistics, the comments of P.V. Sukhatme on social issues and the personal details and perspective of his research provided by C.R. Rao are especially striking.

Contents:
1. Introduction: Radical probabilism (1991)
2. Valuation and acceptance of scientific hypotheses (1956)
3. Probable knowledge (1968)
4. Probability and the art of judgement (1985)
5. Bayesianism with a human face (1983)
6. Alias Smith and Jones: The testimony of the senses (1987)
7. Conditioning, kinematics, and exchangeability (1988)
8. Preference among preferences (1974)
9. On the interpersonal utility theory (1971)
10. Remarks on interpersonal utility theory (1974)
11. Mises redux (1977)
12. Statistical explanation vs statistical inference(1969)
13. New foundations for Bayesian decision theory (1965)
14. Frameworks for preference (1974)
15. Axiomatizing the logic of decision (1978)
16. A note on the kinematics of preference (1977)

Readership: Statisticians with strong interests in philosophy and decision theories; philosophers with strong interests in statistics and decision theories

"These essays explore a variety of topics, ranging from decision theory and the philosophy of mind to epistemology and scientific methodology, from a probabilistic viewpoint of a sort called subjective (Bruno de Finetti's term), personal (L.J. Savage's) and judgemental (mine) ... ."
"Essay 1 is new; the rest, going back over 35 years, appear here essentially unchanged, with after-thoughts clearly labelled ... ."
The term "radical" in Essay 1 relates to the fact that the familiar form of Bayes Theorem P(H|E)=P(E|H)P(H)/P(E) is commonly used when E is certainly known to be true. If, with Jeffrey, we doubt the attainability of certain knowledge, some reformulation of Bayesian "probability kinematics" may be needed. This is the principal preoccupation of Essays 5 and 7. Those who, like I, have difficulty with the argument here are recommended to read the paper by Diaconis and Zabell in J.Amer.Stat.Soc. 77, 882-829.
Those who accept frequency interpretations of probability will be depressed by the suggestion in Essay 11 that von Mises' interpretation is the only one available.
Philosophers working in this field pay much respect to the published and unpublished work of F.P. Ramsey. When they come across his references to Fisher's likelihood, one may expect considerable clarification to ensue.

Contents:
1. Fundamental concepts of total quality management
2. The basic tools
3. The measurement system
4. Statistical process control
5. Using the tools
6. Applications

Readership: Engineers in manufacturing industries interested in quality improvement. No prior knowledge of statistics is assumed, but some facility with basic mathematics would be helpful

This book was "written by working engineers for working engineers" to provide an accessible introduction to statistically simple but fundamentally important methods that can be useful in industrial quality improvement. The book is ambitious, giving very broad but at times abbreviated coverage. The authors acknowledge that some statisticians may find fault with the short-cuts taken in some of the methods. The book's extended Table of Contents, not shown in the condensed version given here, would make a useful checklist of things for statistical practitioners in industry to know. Most of the material in the book is drawn from other sources. I enjoyed Chapter 6, which is a compilation of ways in which the statistical tools presented in earlier chapters can be applied, for example, in customer service, materials control, maintenance, marketing and sales, human resources, purchasing, safety, etc. However, I was disappointed by what appears to have been hasty proofreading, which resulted in errors ranging from the incorrect spelling of the name of R.A. Fisher on page 462 to incorrect statements of the area under the normal curve on page 408. These kinds of errors, which likely occur elsewhere in the book as well, could result in invalid analyses per-formed by non-statisticians who must depend on the absolute correctness of their statistical reference books.

Contents:
PART I : Fundamental Concepts and Methods
PART II : Process Control and Improvement
PART III: Product/Process Design and Improvement

Readership: Engineers in manufacturing industries, and students just beginning their studies in these fields

This book is an elementary treatment of quality control and experimental design, aimed at non-statistical industrial practitioners and beginning students. Although "Design" comes first in its title, a far greater portion of the book is devoted to control charting. Deming's quality management philosophy and some of Taguchi's principles of product design are used to define a conceptual framework for, and to motivate, the methods presented in the book. The selection of topics is sensible. The approach to experimental design taken in the book is largely that of Box and Hunter. Discussion of the mathematical underpinnings of most of the methods is largely omitted, however, which limits the book's usefulness as a stand-alone text in academic departments. Nevertheless the book provides a useful starting point.

Contents:
1. Preliminaries
2. Statistical inference
3. Sampling with replacement
4. Bivariate Poisson distribution
5. Bivariate negative binomial distribution
6. Sampling from finite populations
7. Bivariate logarithmetic series distribution
8. Compound bivariate Poisson distributions
9. Some miscellaneous results

Readership: Graduate students, pure and applied statisticians, researchers in discrete distribution theory

This welcome addition to Dekker's series of monographs on particular distributions begins with a systematic presentation of analytical tools for handling multivariate discrete distributions. These include the use of generating functions, convolutions, various ways of combining distributions, polynomial expansions, and computer simulation procedures.
Most discrete multidimensional distributions are homogeneous in the sense that their probability generating functions have the form G(z1,...zk)=H(a1z1+...+akzk). A number of bivariate discrete distributions lack this property, however. The tools that are provided in the first two chapters are used in later chapters to examine the structural properties of both homogeneous and inhomogeneous bivariate discrete distributions. Each of Chapters 3 through 8 examines bivariate generalizations of a different univariate distribution (binomial, Poisson, negative binomial, hypergeometric, logarithmic, and compound Poisson). Chapter 9 deals rapidly with the following distributions: bivariate Waring; bivariate "short"; bivariate generalized power series; and mixtures of bivariate generalized power series. The arrangement of material within each of these chapters brings out the interrelationships between distributions in different chapters whilst keeping each chapter independent of the others.
The presentation of the material is clear and well-organized, with good lay-outs of formulae. There are over forty pages of references to the somewhat scattered literature, a key word index to the references, and a subject index.

Contents:
1. Introduction
2. Markovian dynamics
3. Discrete probability spaces
4. Independent identically distributed random variables
5. Statistics
6. Markov processes
7. Elements of information theory
8. Fluctuation theory
9. Optimal strategies in casinos: Red and black
10. Foundational problems

'Discrete stochastics' may be translated as: probability and stochastic process theory for discrete distributions and random variables. The author introduces those ideas from measure theory relevant to discrete distributions and builds a simplified probability theory for that case. In this probability framework he discusses some basic stochastic processes, essentially Markov chains and independent, identically distributed random variables, as well as gambling strategies and information theory. As the author points out, there are
penalties for excluding more general probability spaces one of which is that countable sequences of random variables cannot be dealt with. This, in turn, circumscribes the extent to which limit theorems can be covered. The few exercises that are given are woven into the text with the objective, it would appear, more of abbreviating the presentation than as training examples to help reinforce earlier material. There is a tendency to focus on a particular aspect of a topic without giving a sufficiently broad overview that enables the reader to see where the subject is going and where the material fits in. The treatment requires a fair level of mathematical sophistication; explanations are sometimes difficult to follow. In spite of these reservations much of the material covered in the final four chapters is not widely available at an introductory level; thus to this extent the book plays a useful role.

Contents:
1. Probability and probability models
2. Estimation of parameters
3. Hypothesis testing
4. Other selected topics

This book is intended as a second course in statistical methods for biostatisticians. Much of the material is similar to that presented in standard statistical texts, except that examples and many of the problems are based on examples from the biomedical literature. This book is limited in scope, with Chapter 1 covering basic distributions and models, Chapters 2 and 3 focusing on maximum likelihood methods, and Chap-ter 4 covering several unrelated topics. However, the material that is covered appears to be presented clear-ly. This text could serve as a useful text for quantitative epidemiologists or for biostatisticians who do not plan to take an advanced course in statistical inference.

Contents:
1. Contingency tables and the chi-square test
2. 2 x 2 contingency tables
3. r x c contingency tables
4. Multidimensional tables
5. Log-linear models for contingency tables
6. Linear-logistic models
7. Contingency tables with ordered categories
8. Some special types of contingency table

Methods for the analysis of count data have been one of the major developments in statistics over the last fifteen years or so. This is reflected in the second edition of Everitt's book, almost two-thirds of which is devoted to a lucid discussion of the many new techniques available for the analysis of contingency tables of all shapes, sizes and dimensions. Here the reader will find, inter alia, non-technical descriptions of: the graphical display of the information in
a contingency table using correspondence analysis; the fitting and selection of log-linear models; logistic models for dichotomous or polychotomous responses; conditional logistic regression for case control studies; modelling tables with ordered categories and special techniques for incomplete and square tables. For the most part the methods are illustrated with numerical examples and concluded with helpful discussions of issues that arise in their implementation and interpretation.
The first edition of this book was a useful reference for anyone, statistician or otherwise, contemplating the analysis of count data; the second edit-ion is even more so. Now that many of these techniques can be easily implemented with the aid of a package, even by the statistically naive, this book will do much to deepen understanding. The professional statistician will find this book useful too, since it contains good discussions of issues that have been controversial, such as whether or not to apply Yates's correction and what to do when the expected frequencies are small.
All in all, this book is a welcome addition to the fast-growing literature on the analysis of count data.

Contents:
1. First steps
2. One or two samples
3. Multiple comparisons
4. Straight-line data
5. Multiple regression
6. Two-way ANOVA
7. Categorical data
8. Epidemiologic data
9. Multivariate data I
10. Multivariate data II
Two DOS disks with SPIDA program and data sets

This book and the accompanying software and data sets provide the materials for a comprehensive introduction to interactive data analysis, suitable for undergraduates in statistics, graduate students in applied statistics, and users of statistics in other disciplines. The emphasis is on using the tools to discover structure and meaning in data rather than on mathematical formulas and statistical properties.
Installation of the DOS computer package SPIDA (Statistical Package for Interactive Data Analysis) and data sets took me fifteen minutes from first opening of the book; exploring the package's data handling facilities to the end of Chapter 1, another fifteen. From there, I proceeded chapter by chapter, trying the computer examples and following a fascinating exposition of classical statistical procedures as data analysis tools enhanced by techniques of exploratory data analysis. Some not-so-elementary methods are covered, including log-linear models, proportional-hazards regression, and multidimensional scaling. I found the exposition clear and helpful, and used SPIDA with ease.
The eighty-six sets of data, almost all from Data by D.F. Andrews and A.M. Herzerg [Short Book Reviews, Vol. 5, p.38], are sufficiently rich and varied to supply motivation and adequate scope for the "detective work" of interactive data analysis.

Contents:
1. Normal theory models and some extensions
2. Markov chains
3. Point and renewal processes
4. Survival curves
5. Growth curves
6. Time series: The time domain
7. Time series: The frequency domain
8. Repeated measurements
9. Stochastic processes and generalized linear models

The GLIM package is not the obvious choice of a software package to use for the analysis of stochastic processes. However, this book presents clearly how the theory of generalized linear models from which GLIM was originally developed can be adapted to problems of probabilistic modelling. In some cases, the theory must be inferred from a knowledge and understanding of the GLIM codes presented in the text.
The book contains numerous macros written for GLIM 3.77 and is lavishly illustrated with applications. Many of the macros included in the text were originally written by various authors. They have been modified to make them more accessible to the reader not familiar with the details of GLIM programming.
To my mind, the strength of the book and that of the GLIM package lies in the area of preliminary data analysis. Recently, I found Lindsey's treatment of Markov chains and time series to be of considerable utility prior to the development of Fortran codes and the programming of confirmatory analyses using the statistical package BMDP.

Contents:
1. Simple linear regression
2. Detection and correction of model violations: Simple linear regression
3. Multiple regression model
4. Qualitative variables as regressors
5. Weighted least squares
6. The problem of correlated errors
7. Analysis of collinear data
8. Biased estimation of regression coefficients
9. Selection of variables in a regression equation
10. Selected problems

The first edition (1977) had 228 pages; thus this revision has been increased by fifty pages, or approximately twenty-two per cent. The chapter headings are the same as before. The main changes are: (a) an infiltration of newer material on regression diagnostics throughout; (b) an elaboration of work on logistic regression in Chapter 5, of time series in Chapter 6, and of multicollinearity in Chapters 7 and 8; (c) two pages on model fitting strategy in Chapter 9; and (d) a new section of eleven pages called "Selected Prob-lems" containing "seven real data sets with directed questions."
Overall, the book retains its character as a useful series of examples illustrating regression techniques.

Contents:
1. Variable distributions
2. Bivariate regression analysis
3. Basics of multiple regression
4. Regression criticism
5. Fitting curves
6. Robust regression
7. Logit regression
8. Principal components and factor analysis

The subtitle "A Second Course in Applied Statistics" has the implication that the book is de-signed for a wide audience with limited prerequisites. A great variety of material has been packed into relatively few pages in a flowing and discursive way. The exercises contain a variety of examples of real data. There are plenty of diagrams. The mathematical level is comparatively low. The book places "more emphasis on practical issues and troubleshooting than on statistical theory". Students will likely emerge with a wide, but relatively shallow, knowledge of regression. For such an objective, the book is excellent.

Contents:
1. Introduction
2. The logic of hierarchical linear models
3. Principles of estimation and hypothesis testing for hierarchical linear models
4. An illustration
5. Applications in organizational research
6. Applications to the study of individual change
7. Applications in meta-analysis and other cases where level-1 variances are known
8. Three-level models
9. Assessing the adequacy of hierarchical models
10. Technical appendix

Readership: Applied statisticians in all fields, social science research workers

This is a first-class book dealing with one of the most important areas of current research in applied statistics. The authors work in education and the many illustrative examples are drawn from this field, but the methods they describe are as widely applicable as ordinary linear modelling. The standard of exposition is extremely high, though the notation is inevitably complex and non-mathematical research workers may find it tough going. The underlying theory is passed over fairly lightly, with a largely Bayesian approach being adopted. The would-be user of the techniques will need to come to terms with one of the available computer packages, but with this proviso the book shows that many of the most awkward problems of data analysis when there is more than one error stratum are now capable of solution.

Contents:
PART I : The Theory of Correspondence Analysis
PART II : Numerical Example of Correspondence Analysis Exercise Illustrating the Theory
PART III: Reading and Illustration of the Listings
PART IV: Analysis and Interpretation
PART V: Cluster Analysis: Agglomerative Hierarchical Clustering

Readership: Users of data analysis, implicitly ranging from casual users of statistics to research statisticians

At last a text in English on correspondence analysis by perhaps its leading exponent, with three chapters on cluster analysis thrown in for good measure. Parts I, II, III consisting of 244, 63 and 67 pages, respectively cover all the underlying theory, a detailed analysis by hand of a small example (12 observations), and information on how to use a computer to undertake, and interpret output from, correspondence analysis, in the context of two computer programs, a FORTRAN listing being given for one. It is suggested that these three parts are read in parallel, rather than in sequence, with a selection of chapters depending on the reader's background. Part IV (183 pages) consists mainly of the detailed analysis of sixteen examples, illustrating the wide-ranging applications of correspondence analysis. Various general practical points are made in an introductory section to this part. Part V (83 pages) concentrates on one particular variety of agglomerative hierarchical clustering, with a distance measure based on inertia. Each part starts with a useful summary. The index also acts as a glossary.

Contents:
1. General introduction
2. Likelihood-based approaches to discrimination
3. Discrimination via normal models
4. Distributional results for discrimination via normal models
5. Some practical aspects and variants of normal theory-based discriminant rules
6. Data analytic considerations with normal theory based discriminant analysis
7. Parametric discrimination via non-normal models
8. Logistic discrimination
9. Nonparametric discrimination
10. Estimation of error rates
11. Assessing the reliability of the estimated posterior probabilities of group membership
12. Selection of feature variables in discriminant analysis
13. Statistical image analysis

Readership: Applied and theoretical statisticians and researchers making use of discriminant analysis

This is an impressive volume. It provides a comprehensive and deep account of the state of the art of discriminant analysis. The extent of coverage is indicated by the fact that it contains over 1200 references. The depth is indicated by the chapter headings. It is sufficiently up to date to include detailed discussion in Chapter 9 of kernel methods, nearest neighbour methods, tree based methods and other non-parametric methods. Mention of appropriate software packages appears at the relevant places in the text.
The author describes the book as 'a monograph, not a textbook' and it does not contain exercises. It does not shy from the necessary mathematics but is nevertheless written in an accessible style.
It seems almost churlish to cavil at things inadequately dealt with in the face of such a tour de force, but I would have liked to see some discussion of choice of technique and also more than the brief mention on page 21 of alternatives to error rate as comparison criteria. Brier score, for example, does not even appear in the index. In general, however, I find it difficult to think what more one could ask of a book devoted to discriminant analysis. This will be the standard work in the area for some years to come. Any statistician who has occasion to use discriminant analysis, which surely must include most statisticians, should buy a copy.

Contents:
1. Statistics for spatial data
PART I : Geostatistical Data
2. Geostatistics
3. Spatial prediction and kriging
4. Applications of geostatistics
5. Special topics in statistics for spatial data
PART II : Lattice Data
6. Spatial models on lattices
7. Inference for lattice models
PART III: Spatial Patterns
8. Spatial point patterns
9. Modeling objects

This is an enormous volume, which is really too large to read comfortably yet not well structured for reference use. Much of it reads as a catalogue of work in spatial statistics without the comments and comparisons which would give the work structure and added value. Where there are comparisons and summaries, I frequently found myself in disagreement, especially as implied criticisms often contradict the chronological order!
A number of examples are interspersed through-out the text, but with very few exceptions these are taken from earlier publications of the author and illustrate his own methodological contributions. Thus there are rather few examples of the standard methods in the field. Because of this, newcomers to the field may gain a misleading impression and would be better advised to read one of the now standard texts of Ripley [Short Book Reviews, Vol. 2 p.5 and Vol. 9. p.7], Diggle [Short Book Reviews, Vol. 4. p.38], Upton and Fingleton [Short Book Reviews, Vol. 5. p.41 and Vol. 9. p.45]. Cressie's book is a useful additional reference, especially to his own work which is widely scattered through the scientific literature.

Contents:
1. Introduction
2. General concepts
3. General results for balanced designs
4. The one-fold nested design
5. The two-fold and (Q-1)-fold nested design
6. Crossed random designs
7. Mixed models

The book gives a careful account of methods for calculating confidence limits for variance components, linear combinations of variance components and ratios of variance components. The general organization of the book will be clear from the list of chapter titles. For those cases in which no "standard" exact treatment is available, the various suggestions in the literature, for example those based on Fairfield Smith, Welch, Satterthwaite-type chi-squared approximations, are described. Numerical examples and brief SAS pro-grams are included.
The book is restricted to linear models based on normal theory. A general criticism is that, while the exposition is admirably clear, the methods tend to appear as a series of ad hoc tricks. Even in the one-way unbalanced case, while a number of methods are well described and the literature on their comparison summarized, one is left without a clear view of why one method works better than another. There is no systematic use of likelihood-based arguments, for example.

Contents:
1. Introduction
2. History and comment
3. The 1-way classification
4. Balanced data
5. Analysis of variance estimation for unbalanced data
6. Maximum likelihood and restricted maximum likelihood
7. Prediction of random variables
8. Computing maximum likelihood and restricted maximum likelihood estimates
9. Hierarchical models and Bayesian estimation
10. Binary and discrete data
11. Other procedures
12. The dispersion-mean model

This substantial work of scholarship is likely to be a major source of information on variance components for a good many years to come. The book proceeds at a fairly leisurely pace from a careful discussion of the balanced one-way classification, through to a discussion of maximum likelihood and restricted maximum likelihood estimation in generality. The final chapters deal with more advanced issues, including a relatively short chapter on binary and discrete data.
Inevitably and properly the emphasis and contents reflect both the authors' special interests and more broadly the substantial body of work to emerge from Cornell over the years. Thus Henderson's various
methods are described in some detail, even though they have to an appreciable extent been replaced by the general availability of programs for restricted maximum likelihood estimation. Matrix formulations, often superficially rather formidable, are adopted at relatively early stages of the treatment. While some simple numerical examples are given they are quite formal and are not intended to give insight into how variance components are used in applications.
Criticisms that might be made of the book include the relatively light emphasis on generalized linear models, the absence of information on the back-ground to examples and the somewhat associated failure to study slightly nonstandard questions attached to the classical well-loved models or how to look for departures from these models.

Contents:
1. Designs
2. Codes
3. The geometry of vector spaces
4. Symmetric designs
5. The standard geometric codes
6. Codes from planes
7. Hadamard designs
8. Steiner systems

The connection between design theory and coding theory has long been one of the most interesting and exciting areas of combinatorics, and a book that gives an account of the linkage has long been needed. The authors present general material in the first five chapters, including a new approach to Reed-Muller codes, and then discuss applications of coding theory to some of the more important classes of designs, namely, finite planes, Hadamard designs, and Steiner systems. The bibliography is long, very complete, and very up-to-date, including references to preprints. The book provides a very valuable research reference for workers in both design theory and coding theory.

Contents:
1. Preliminaries
2. Quadratic forms in real variables
3. Quadratic forms in random variables
4. The distribution of quadratic forms
5. Chi-squaredness and independence
6. Generalized quadratic forms
7. Applications

Readership: Undergraduate and postgraduate students in mathematical statistics, theoretical and applied statisticians

Quadratic forms in random variables arise in many branches of statistics and the authors have provided a valuable comprehensive source of their distributional properties in normal and non-normal cases. Illustrations of their applicability in goodness-of-
fit tests, linear models and variance components, multivariate analysis, stochastic processes and time series analysis, coverage and optimal control problems, incomplete block designs, multivariate functional relationship models and optimization form the final chapter of the book. The early chapters of the book are devoted to a useful review of necessary matrix algebra and results for quadratic forms in real variables, making the contents, at least in part, readily available to undergraduate students in mathematical statistics.

Contents:
PART 1: Introduction
PART 2: General Principles of the Bootstrap
PART 3: Applications of the Bootstrap

This book, based on papers presented at a special topics meeting of the Institute of Mathematical Statistics held in May 1990, describes recent developments of Efron's bootstrap. A readable introduction and overview by Efron and LePage is followed by a series of fifteen fairly technical papers on theoretical and methodological issues, including consistency of the bootstrap in general settings, efficient computation, higher-order refinement and applicability of the boot-strap to dependent data, including Markov processes and time series. The book concludes with a series of eight papers dealing with applications of the bootstrap to areas such as exploratory regression, model select-ion, survival analysis and genetics. The volume represents much of current thinking on the bootstrap; be-cause it is a conference proceedings, there are a number of notable absentees from the list of authors. It is a valuable collection of research papers.

Contents:
1. Introduction
2. The autocorrelation methods
3. Penalty function methods
4. Innovation regression methods
5. Pattern identification methods
6. Testing hypothesis methods

The aim of this book is to unify most of the significant work on identifying autoregressive moving average models. The methods discussed include final prediction error, Akaike information criterion, Bayesian information criterion, inverse autocorrelation functions, and pattern identification methods such as the Corner method. The book is written in a very read-able style, concentrating its effort on the most important characteristics of the techniques covered, and giving extensive references for those interested in more details. One of its strengths is that the book nicely bridges the gap between the statistical and electrical engineering/signal processing literatures;
results from the latter are well represented. It would have been interesting to have seen more discussion of order determination of autoregression processes (by, say final prediction error) when parameter estimation is by techniques other than Yule-Walker (for example, by Burg's method). Also, the book would have benefited from some illustrations. These are, however, minor criticisms. I liked the book very much, and think it will prove to be very useful.

Contents:
Introduction
1. The measure-theoretic prediction process
2. Topological considerations and prediction space: Application to Markov processes
3. Gaussian processes and the prediction process in the wide sense
4. A classification of measurable processes
5. Prediction of measurable processes
6. Application to concrete examples, and ramifications of the theory

This book is an account of how the mathematical theory of continuous time stochastic processes might have developed. Two processes are considered which co-exist on the same probability space, one re-presenting stationary physical time, the other moving observer's time, and together they make the observer's process strongly Markovian. This leads to a representation of any measurable stochastic process in terms of martingales, and the mechanisms for their prediction.
This is given a careful mathematical treatment involving some hard mathematics. Whilst it may be too late for seasoned probabilists to embrace these ideas wholeheartedly, those embarking on their careers as probabilists will find this strongly argued thesis, stimulating and worth working at.
The writing and the production are both at a very high standard. Readers need a basic knowledge and interest in measure theory, functional analysis, stochastic integration, and martingale theory.

This is the first paperback edition of a book published in 1990. Its aim is to give a systematic exposition of the analysis of transition data from an econometric point of view but it enriches it with many insights taken from the statistical survival literature.
The author stresses that the subject of the book is the analysis of transition data and not only of duration data. The destination at the end of a spell
is indeed of interest in many, not only econometric, applications. After pointing at the different jargon used in econometric and biostatistic literature, the author discusses in Part I models for single spell (alias survival) data with emphasis given to the case of neglected heterogeneity (frailty) and of time-varying endogenous (internal) covariates. Further multiple destinations and multiple cycles (competing risks and multi-state) models are discussed in detail.
Part II deals with the inferential problems encountered when fitting parametric or semi-parametric models. It suggests graphical and numerical methods to detect misspecification of the assumed model. The methods proposed for multi-state models are not familiar to biometricians and are therefore particularly interesting.
This is a very good book in its own right but also since it succeeds in linking the econometric and biometric literature on the subject.

Contents:
1. The what, why, and how of wavelets
2. The continuous wavelet transform
3. Discrete wavelet transforms: Frames
4. Time-frequency density and orthonormal bases
5. Orthonormal bases of wavelets and multiresolution analysis
6. Orthonormal bases of compactly supported wavelets
7. More about the regularity of compactly supported wavelets
8. Symmetry for compactly supported wavelets
9. Characterization fo functional spaces by means of wavelets
10. Generalizations and tricks for orthonormal wavelet bases

Readership: Functional analysts, mathematicians, pure and applied scientists, anyone with an interest in wavelets

This compendium is as much a thorough introduction to wavelets as it is a history of the subject and the author's work in the field. This fully referenced and annotated series of lectures was presented at the CBMS-NSF conference on wavelets in June, 1990. The volume reads much more like a book than a lecture series and anyone with an interest in wavelets that is looking for a starter, or a good reference of Daubechies' work all-under-one-roof, will not be disappointed.
Wavelets are families of functions constructed by the dilation and translation of a function with suf-ficient decay in time and frequency. This work is directed predominantly at the graduate level mathematician; hence a reasonable knowledge of mathematics is essential for thorough understanding. It is still possible, however, to follow much of the development with less rigour by simply accepting some of the relation-ships as said. The beauty of this work is the time and patience with which all of the material is laid out. From the "what, why, and how" of Chapter 1, through the definitions of the continuous and discrete wavelets
including new results on the windowed Fourier transform, on the details of construction and regularity, the style remains remarkably readable despite thorough groundwork and careful attention to detail. An inter-change of carefully derived theorems with healthy doses of hand-waving and personalized analysis are complemented by many computed values and figures, many examples, and the inclusion of a number of useful tricks.
This work should serve well both as an introduction to the uninitiated and as a valuable reference book in any library. The statistician might justify the purchase after a perusal of a curious result in the last paragraph of page ninety-eight.

Contents:
1. Monte Carlo methods and quasi-Monte Carlo methods
2. Quasi-Monte Carlo methods for numerical integration
3. Low-discrepancy point sets and sequences
4. Nets and (t,s)-sequences
5. Lattice rules for numerical integration
6. Quasi-Monte Carlo methods for optimization
7. Random numbers and pseudorandom numbers
8. Nonlinear congruential pseudorandom numbers
9. Shift-register pseudorandom numbers
10. Pseudorandom vector generation

Readership: Statisticians interested in theoretical properties of simulation methods, computer scientists

A quasi-Monte Carlo method can be described as the deterministic version of a Monte Carlo method, where the random samples of the Monte Carlo method are replaced by deterministic points, selected to produce a deterministic error bound smaller than the probabilistic Monte Carlo error bound. This monograph is an expanded account of a series of talks given by the author at a conference held at the University of Alaska in August 1990. Useful background material is included, though the bulk of the monograph is taken up with a very technical and dense account of the theoretical properties of quasi-Monte Carlo methods. The treatment is very full but clear. The statistical practitioner is likely to find more interest in the later chapters which describe recent advances in pseudorandom number generation. Quasi-Monte Carlo methods underlie the theoretical analysis of the various methods for the generation of uniform pseudorandom numbers discussed. An interesting monograph rather than a reference work.

Readership: Graduate students in computer science, algebra, logic and related areas

Modern universal algebra arose in the 1930's from studying general algebraic constructions in equationally defined systems like semigroups, lattices, Boolean algebras, groups, rings, etc.
The book under review is a very well-written graduate course in universal algebra inspired by applications in theoretical computer science. Chapters 1 and 2 give a thorough treatment of term rewriting theory (some topics: structural induction, fixed-point theory, rewriting systems, termination orderings, word problems for congruences). Chapter 3 presents varieties and quasi-varieties of algebras and their model theory (some topics: free algebras, representation theorems, Horn and conditionally equational theories). Chapter 4 gives applications to semantics of programming languages and algebraic data types. In an unusual step, the author treats the important topics of multi-sorted and order-sorted, as well as w-complete algebras.
The book is very clear and makes a good and interesting introduction to many aspects of universal algebra, while stressing applications to computer science and formal languages. Unfortunately, there are no exercises, which somewhat impedes its pedagogical aims.

Readership: Researchers in stochastic analysis, probabilists and mathematical statisticians

In 1938, N. Wiener introduced the fact that any square-integrable function of Brownian motion is representable as a series (over n) of n-th order multiple integrals with respect to product Brownian motion. To statisticians, this result could be re-phrased in terms of every finite-variance statistic that is a function of Brownian motion being expressible as a series of U-statistics of increasing degree. From the 1951 formulation of this result by K.Ito to the present, there has been extensive work done on these "polynomial chaoses", the expression by Wiener for the terms of the series.
In this text an excellent presentation is made of the by now extensive literature on random series and stochastic integrals, both single and multiple. Particularly fine expositions of hypercontractivity, polynomial chaos and independently scattered random measures are included, all being key topics in the back-ground for the book's main emphasis. The book also pulls together a lot of material including probability inequalities that will be of interest to a wider group of probabilists and statisticians than suggested by the book's main coverage.

Contents:
1. Introduction
2. Identifiability of distributions of random variables based on some functions of them
3. Identifiability of probability measures on abstract spaces
4. Identifiability for some types of stochastic processes
5. Generalized convolutions
6. Identifiability in some econometric models
7. Identifiability in reliability and survival analysis
8. Identifiability for mixtures of distributions

This book carries as a subtitle 'Characterisation of Probability Distributions'. In Chapter 2, the following problem is discussed: let X1,X2,X3 be three independent real-valued random variables. Define Z1=X1-X3,Z2=X2-X3. Under which condition on (Z1,Z2), does the joint distribution of (Z1,Z2) determine the distribution of X1,X2,X3 up to a change of location. Chapters 2, 3 and 4 all deal with variants of this problem. After a brief excursion on generalized con-volutions, Chapters 6, 7 and 8 all deal with the problem of statistical identifiability, both in the para-metric as well as in the non-parametric case. Though some of the problems treated have important practical applications, the style in which the book is written emphasizes much more the mathematical methodology underlying the analytic solutions. The various theorems given are often complemented by specific, though in many cases artificial, examples. Ploughing through the text the reader has to be prepared to jump from the real line, over Polish spaces to topological groups and semigroups having visited also Stiefel manifolds on the way. The specialist in the field might perhaps welcome this text, I very much doubt, however, that the more applied probabilist or statistician will ge much out of it.

Contents:
1. Foundations of stochastic approximation, H. Walk
2. Applicational aspects of stochastic approximation, G. Pflug
3. Applications to adaptation algorithms, L. Ljung

It is pleasant to find hard results collected together in such an elegant fashion. For those interested in the frontier between statistics and control theory this monograph is a must. The increasing use of stochastic approximation in non-linear control is a pleasant surprise to those who bemoan the separation between the two fields. It should also be useful to research workers in stochastic learning and related fields.

Contents:
1. Linear filtering theory
2. Optimal stochastic control for linear dynamic systems with quadratic payoff
3. Optimal stochastic control of linear stochastic systems with an exponential-of-integral performance index
4. Nonlinear filtering theory
5. Perturbation methods in nonlinear filtering
6. Some explicit solutions of the Zakai equation
7. Some explicit controls for systems with partial observation
8. Stochastic maximum principle and dynamic programming for systems with partial observation
9. Existence results for stochastic control problems with partial information

This research monograph gives a comprehensive survey of optimal filtering theory and optimal control for partially observed stochastic systems. The early part of the book deals with cases where finite dimensional statistics exist, for example, linear problems with quadratic type loss functions. Later in the book more general problems are studied where the solution is infinite dimensional. Bensoussan is well known for his contributions to these areas and this book provides a lucid account of his results plus the key background results of other key figures in the development of stochastic control theory (Kalman, Bucy, Fleming, Wonham, Whittle, Kushner, Zakai etc.). The book is recommended to those who would like a succinct introduction to this important topic.

Title	SCIENCE, IDEOLOGY, AND THE MEDIA. The Cyril Burt Scandal.
Author	R. Fletcher.
Publisher	New Brunswick, New Jersey: Transection Publishers, 1991, pp. xxviii + 419, US$ 29.95.

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

Title	MEDICAL USES OF STATISTICS, 2nd edition.
Author	J.C. Bailar III and F. Mosteller (Eds.).
Publisher	Boston: NEJM Books, 1992, pp. xxvii + 449, US$39.95.

Reviewer:
Institute	York University
Place	North York, Canada
Name	P.A. Herzberg

Title	MACSYMA FOR STATISTICIANS.
Author	B. Heller.
Publisher	New York: Wiley, 1991, pp. xiv + 246, ,32.15.

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

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

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

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

Reviewer:
Institute	Imperial Cancer Research Fund
Place	London, U.K.
Name	B.L. De Stavola

Reviewer:
Institute	University of Southampton
Place	Southampton, U.K.
Name	T.M.F. Smith

Reviewer:
Institute	University of Auckland
Place	Auckland, New Zealand
Name	A.J. Scott

Reviewer:
Institute	Statistics Canada
Place	Ottawa, Canada
Name	C.A. Mills

Reviewer:
Institute	London School of Economics and Political Science
Place	London, U.K.
Name	C. O'Muircheartaigh

Reviewer:
Institute	Imperial College of Science, Technology and Medicine
Place	London, U.K.
Name	L.V. White

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

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

Reviewer:
Institute	University of Leeds
Place	Leeds, U.K.
Name	K.V. Mardia

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

Reviewer:
Institute	University of Oxford
Place	Oxford, U.K.
Name	A.C. Davison

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

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

Reviewer:
Institute	ETH-Zürich
Place	Zürich, Switzerland
Name	A.S. Sznitman

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

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

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

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

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

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

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

Reviewer:
Institute	Museu Paraense Emilo Goeldi
Place	Belem, Para, Brazil
Name	C.M. O'Brien

Reviewer:
Institute	Institute of Education, University of London
Place	London, U.K.
Name	M.J.R. Healy

Reviewer:
Institute	University of Aberdeen
Place	Aberdeen, U.K.
Name	I.T. Jolliffe

Reviewer:
Institute	University of Manitoba
Place	Winnipeg, Canada
Name	R.G. Stanton

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

Reviewer:
Institute	Andersen Consulting
Place	Toronto, Canada
Name	J.R. Whitla