Short Book Reviews On-Line 1993

ALGORITHMS, ROUTINES, AND S FUNCTIONS FOR ROBUST STATISTICS. THE FORTRAN LIBRARY ROBETH WITH AN INTERFACE TO S-PLUS.	A. Marazzi. With the collaboration of J. Joss and A. Randriamiharisoa.
INTRODUCTION TO SHAPE OPTIMIZATION. Shape Sensitivity Analysis.	J. Sokolowski and J.-P. Zolesio.
NUMERICAL METHODS FOR STOCHASTIC CONTROL PROBLEMS IN CONTINUOUS TIME.	H.J. Kushner and P.G. Dupuis.
DYNAMIC PROGRAMMING.	M. Sniedovich.
SENSE AND NONSENSE OF STATISTICAL INFERENCE. Controversy Misuse and Subtlety.	C. Wang.
UNIVARIATE DISCRETE DISTRIBUTIONS, 2nd edition.	N.L. Johnson, S. Kotz and A.W. Kemp.
A HANDBOOK OF GENERALIZED SPECIAL FUNCTIONS FOR STATISTICAL AND PHYSICAL SCIENCES.	A.M. Mathai.
STATISTICAL PROCESS CONTROL FOR QUALITY IMPROVEMENT.	J.R. Thompson and J. Koronacki.
SAMPLING.	S.K. Thompson.
NONSAMPLING ERROR IN SURVEYS.	J.T. Lessler and W.D. Kalsbeek.
SURVEY SAMPLING. Theory and Methods.	A. Chaudhuri and H. Stenger.
LONGITUDINAL DATA WITH SERIAL CORRELATION: A STATE-SPACE APPROACH.	R.H. Jones.
THE DESIGN AND ANALYSIS OF RESEARCH STUDIES.	B.F.J. Manly.
QUESTIONS ABOUT QUESTIONS: INQUIRIES INTO COGNITIVE BASES OF SURVEYS.	M. Tanur (Ed.).
STATISTICS IN THE ENVIRONMENTAL AND EARTH SCIENCES.	A.T. Walden and P. Guttorp (Eds.).
CLUSTER ANALYSIS, 3rd edition.	B.S. Everitt.
OPTIMUM EXPERIMENTAL DESIGNS.	A.C. Atkinson and A.N. Donev.
DATA FITTING IN THE CHEMICAL SCIENCES.	P. Gans.
BIOASSAY, 3rd edition.	J.J. Hubert.
CROSS-OVER TRIALS IN CLINICAL RESEARCH.	S. Senn.
CROSS-OVER EXPERIMENTS: DESIGN, ANALYSIS AND APPLICATION.	D.A. Ratkowsky, M.A. Evans and J.R. Alldredge.
STATISTICS AND DECISIONS.	S.H. Kim.
PROBABILITY THEORY. An Introductory Course.	Y.G. Sinai. Translated from the Russian by D. Haughton.
RESAMPLING-BASED MULTIPLE TESTING: Examples and Methods for p-Value Adjustment.	P.H. Westfall and S.S. Young.
WHEN DOES BOOTSTRAP WORK? ASYMPTOTIC RESULTS AND SIMULATIONS.	E. Mammen.
INFORMATION BOUNDS AND NONPARAMETRIC MAXIMUM LIKELIHOOD ESTIMATION.	P. Groeneboom and J.A. Wellner.
SPATIAL TESSELLATIONS CONCEPTS AND APPLICATIONS OF VORONOI DIAGRAMS.	A. Okabe, B. Boots and K. Sugihara. With a foreword by D.G. Kendall.
RANDOM FIELD MODELS IN EARTH SCIENCES.	G. Christakos.
STATISTICAL MODELS BASED ON COUNTING PROCESSES.	P.K. Andersen, O. Borgan, R.V. Gill and N. Keiding.
STOCHASTIC EQUATIONS IN INFINITE DIMENSIONS.	G. Da Prato and J. Zabczyk.
WEIGHTED APPROXIMATIONS IN PROBABILITY AND STATISTICS.	M. Csörgö and L. Horváth.
AN INTRODUCTION TO PROBABILITY AND STOCHASTIC PROCESSES.	M.A. Berger.
STOPPING TIMES AND DIRECTED PROCESSES. G.A. Edgar and	L. Sucheston.
EXCURSIONS OF MARKOV PROCESSES.	R.M. Blumenthal.
REGENERATIVE STOCHASTIC SIMULATION.	G.S. Shedler.
POISSON PROCESSES.	J.F.C. Kingman.
CHAMPS ALÉATOIRES SUR UN RÉSEAU.	X. Guyon.
ARTIFICIAL INTELLIGENCE FRONTIERS IN STATISTICS: AI AND STATISTICS III.	D.J. Hand (Ed.).
SIMULATION: A STATISTICAL PERSPECTIVE.	J.P.C. Kleijnen and W. van Groenendaal.
REPRESENTATION AND CONTROL OF INFINITE DIMENSIONAL SYSTEMS. Volume 1	A. Bensoussan, G. Da Prato, M.C. Belfour and S.K. Mitter.
MARKOV DECISION PROCESSES.	D.J. White.
NETWORK MODELS IN OPTIMIZATION AND THEIR APPLICATIONS IN PRACTICE.	F. Glover, D. Klingman and N.V. Phillips.
CHAOTIC AND FRACTAL DYNAMICS. An Introduction for Applied Scientists and Engineers.	F.C. Moon.
THE ANALYTICS OF UNCERTAINTY AND INFORMATION.	J. Hirshleifer and J.G. Riley.
MATHEMATICS IN MEDICINE AND THE LIFE SCIENCES.	F.C. Hoppensteadt and C.S. Peskin.
DESIGN AND ANLAYSIS OF BIOAVAILABILITY AND BIOEQUIVALENCE STUDIES.	S.C. Chow and J.P. Liu.
RISK: ANALYSIS, PERCEPTION AND MANAGEMENT. Report of a Royal Society Study Group	-
PROBABILITY AND STATISTICS IN EXPERIMENTAL PHYSICS.	B.P. Roe.
BIOSTATISTICS. A METHODOLOGY FOR THE HEALTH SCIENCES.	L.D. Fisher and G. van Belle.
PROCESS CAPABILITY INDICES.	S. Kotz and N.L. Johnson.
STATISTICAL METHODS OF QUALITY ASSURANCE.	H.J. Mittag and H. Rinne.
SEQUENTIAL DATA IN BIOLOGICAL EXPERIMENTS. An Introduction for Research Workers.	E.A. Roberts.
STATISTICAL METHODS IN ANALYTICAL CHEMISTRY.	P.C. Meier and R.E.Zünd.
ANALYTICAL POPULATION DYNAMICS.	T. Royama.
STATISTICS IN THEORY AND PRACTICE.	R. Lupton.
A SHORT COURSE IN EPIDEMIOLOGY.	S. Norell.
STATISTICAL METHODS FOR SURVIVAL DATA ANALYSIS, 2nd edition	E.T. Lee.
THE USE OF RESTRICTED SIGNIFICANCE TESTS IN CLINICAL TRIALS.	D.S. Salsburg.
INTRODUCTION TO PROBABILITY AND STATISTICS, 2nd edition, Revised and Expanded.	N.C. Giri.
DATA ANALYSIS FOR COMPARATIVE SOCIAL RESEARCH: INTERNATIONAL PERSPECTIVES.	C. Hayashi, T. Suzuki and M. Sasaki.
VERTICALLY TRANSMITTED DISEASES: MODELS AND DYNAMICS.	S. Busenberg and K. Cooke.
APPLIED FACTOR ANALYSIS IN THE NATURAL SCIENCES.	R.A. Reyment and K.G. Jöreskog. Appendix by L.F. Marcus.
FUNDAMENTALS OF PATTERN RECOGNITION, 2nd edition.	M. Pavel.
MULTIVARIATE PATTERN RECOGNITION IN CHEMOMETRICS.	Illustrated Case Studies. R.G. Brereton (Ed.).
PITMAN'S MEASURE OF CLOSENESS. A COMPARISON OF STATISTICAL ESTIMATORS.	J.P. Keating, R.L. Mason and P.K. Sen.
APPLIED NONPARAMETRIC STATISTICAL METHODS, 2nd edition.	P. Sprent.
DIFFERENTIAL GEOMETRY AND STATISTICS.	M.K. Murray and J.W. Rice.
A COURSE ON POINT PROCESSES.	R.D. Reiss.
DISCRETE EVENT SYSTEMS. Sensitivity Analysis and Stochastic Optimisation by the Score Function Method.	R.Y. Rubinstein and A. Shapiro.
MAXIMUM ENTROPY SOLUTIONS TO SCIENTIFIC PROBLEMS.	R.M. Bevensee.
SPECTRAL ANALYSIS FOR PHYSICAL APPLICATIONS. Multitaper and Conventional Univariate Techniques.	D.B. Percival and A.T. Walden.
BIRTH AND DEATH PROCESSES AND MARKOV CHAINS.	Z. Wang and Y. Yang.
MOMENTS IN PROBABILITY AND APPROXIMATION THEORY.	G.A. Anastassiou.
CONTROLLED MARKOV PROCESSES AND VISCOSITY SOLUTIONS.	W.H. Fleming and H.M. Soner.
THERMODYNAMICS OF CHAOTIC SYSTEMS. An Introduction.	C. Beck and F. Schlögl.
WEIGHTED EMPIRICALS AND LINEAR MODELS.	H.L. Koul.
WAVELETS AND OPERATORS.	Y. Meyer. Translated by D.H. Salinger.
MARTINGALE SPACES AND INEQUALITIES.	R. Long.
CODING AND INFORMATION THEORY.	S. Roman.
INFORMATION THEORY AND MOLECULAR BIOLOGY.	H.P. Yockey.
ANALYSIS OF QUANTAL RESPONSE DATA.	B.J.T. Morgan.
PROBABILITY AND RANDOM PROCESSES, 2nd edition.	G.R. Grimmett and D.R. Stirzaker.
PROBABILITY AND RANDOM PROCESSES. PROBLEMS AND SOLUTIONS.	G.R. Grimmett and D.R. Stirzaker.
PROBABILITY AND ITS APPLICATIONS FOR ENGINEERS.	D.H. Evans.
ADVENTURES IN STOCHASTIC PROCESSES.	S.I. Resnick.
APPLIED MULTIVARIATE DATA ANALYSIS. Volume II: Categorical and Multivariate Methods.	J.D. Jobson.
A FIRST COURSE IN ORDER STATISTICS.	B.C. Arnold, N. Balakrishnan and H.N. Nagaraja.
PREDICTION THEORY FOR FINITE POPULATIONS.	H. Bolfarine and S. Zacks
AN INTRODUCTION TO STOCHASTIC PROCESSES AND THEIR APPLICATIONS.	P. Todorovic.
MULTIVARIATE DENSITY ESTIMATION: THEORY PRACTICE AND VISUALIZATION.	D.W. Scott.
CURRENT ISSUES IN STATISTICAL INFERENCE: ESSAYS IN HONOR OF D. BASU.	M. Ghosh and P.K. Pathak, (Eds.).
ITEM RESPONSE THEORY: PARAMETER ESTIMATION TECHNIQUES.	F.B. Baker.
NUMERICAL SOLUTION OF STOCHASTIC DIFFERENTIAL EQUATIONS	P.E. Kloeden and E. Platen.
QUEUEING AND RELATED MODELS.	U.N. Bhat and I.V. Basawa(Eds.). Foreword by N.U. Prabhu.
LECTURES ON THE COUPLING METHOD.	T. Lindvall.
ANALYSIS OF RANDOM WALKS.	J.W. Cohen.
PASSAGE TIMES FOR MARKOV CHAINS.	R. Syski.
RANDOM WALKS, CRITICAL PHENOMENA, AND TRIVIALITY IN QUANTUM FIELD THEORY.	R. Fernández, J. Fröhlich and A.D. Sokal.
AN INTRODUCTION TO THE MODELLING OF NEURAL NETWORKS.	P. Peretto.
THE DESIGN OF RELATIONAL DATABASES.	H. Mannila and K.J. Räihä.

Contents:
Introduction
1. Location problems
2. M-estimates of coefficients and scale in linear regression
3. Weights for bounded influence regression
4. Covariance matrix of the coefficient estimates
5. Asymptotic relative efficiency of regression M-estimates
6. Robust testing in linear models
7. High breakdown point regression
8. M-estimates of covariance matrices
9. Mixed procedures
10. M-estimates for discrete generalized linear models
11. Weight functions
12. Utility routines
13. FORTRAN sources
14. S-PLUS interface

Readership: Statisticians interested in statistical computing, experimental scientists and research workers interested in modern statistical methods

This book is a collection of algorithms that allows the compution of a broad class of robust statistical procedures. These procedures are based on M-estimation and high breakdown point estimation, including robust regression, robust testing of linear hypotheses and robust covariances. All of the algorithms are implemented into the FORTRAN subroutine library ROBETH. It is possible to use ROBETH routines with many common packages, such as SAS, SPSS, etc., provided tools for interfacing FORTRAN routines exist in these packages. The authors discuss this interface to the statistical environment S-PLUS. The book is well organized and readable. Most chapters are organized into three primary sections: the first describes the statistical methodology, the second contains the description of FORTRAN subroutines and the third presents examples for using subroutines by means of S-PLUS interface. Anyone with an interest in statistical software will find this book and the added software useful.

Contents:
1. Introduction to shape optimization
2. Preliminaries and the material derivative method
3. Shape derivatives for linear problems
4. Shape sensitive analysis of variational inequalities

Shape optimization refers to the problem of adjusting the shape of an object in order to achieve the best performance for a given purpose. For example, one might ask what is the best shape for the fins on a radiator in order to dissipate the most hcat from a specified volume of material. Problems of this nature occur in many areas including thermodynamics, fluid mechanics, structures and electronics. The solution of these problems depends upon variational methods for systems governed by partial differential equations. This book outlines a range of approaches to these problems and shows how appropriate differentials can be obtained for different classes of problems. This re-search monograph is recommended to people who are interested in solving engineering problems involving the optimization of shape.

Contents:
1. Review of continuous time models
2. Controlled Markov chains
3. Dynamic programming equations
4. The Markov chain approximation method: Introduction
5. Construction of the approximating Markov chain
6. Computational methods for controlled Markov chains
7. The ergodic cost problem: Formulations and algorithms
8. Heavy traffic and singular control problems: Examples and Markov chain approximations
9. Weak convergence and the characterization of processes
10. Convergence proofs
11. Convergence for reflecting boundaries, singular control and ergodic cost problems
12. Finite time problems and nonlinear filtering
13. Problems from the calculus of variations
14. The viscosity solution approach to proving convergence of numerical solutions

Readership: Mathematicians with an interest in stochastic processes and stochastic control

This book is concerned with the numerical solution of various problems in the study of stochastic processes. The essential idea is to approximate the problem via a Markov chain on a finite state space. The discretized problem can then be readily solved. This procedure is described in a general format so it can be applied to a wide range of problems. The approximations are developed in such a way that properties of the Markov chain approach those of the original problem as a discretization parameter approaches zero. For those readers who are interested in the formal verification of these facts, the proofs are given based on Weak Convergence Theory. This is an interesting book for anybody who really wants to solve a stochastic control problem. The key point made in the book is that naive discretization methods may simply not work and thus there is strong motivation to study discretization methods with guaranteed convergence properties.

Contents:
1. Introduction
PART I : Science
2. Fundamentals
3. Multistage decision model
4. Dynamic programming: An outline
5. Solving the functional equation
6. Successive approximations method
7. Optimal policies
8. The curse of dimensionality
9. A perspective on Part I
PART II : Art
10. Refinements
11. The state
12. Parametric schemes
13. The principle of optimality
PART III: Epilogue
14. What then is dynamic programming?
APPENDIX A: Contraction Mapping
APPENDIX B: Fractional Programming
APPENDIX C: C-Programming
APPENDIX D: The Principles of Optimality in Stochastic Processes

This text approaches dynamic programming as a methodology, as an approach to problem solving. The discussion is directed to discussing and characterizing the constituents of the dynamic programming model. It uses examples to demonstrate how to make concrete the model's components for specific problems. The use of examples to illustrate the theory avoids the tangle of theory and the idiosyncrasies of a problem that is all too common in dynamic programming texts. The author presents no computational schemes for solving problems; his aim is to elucidate the fundamental and universal character of the dynamic programming model and in this he succeeds completely.

Contents:
1. Fads and fallacies in hypothesis testing
2. Quasi-inferential statistics
3. Statistical causality and law-like relationships
4. Amoeba regression and time series models
Intermission
5. A critical eye and appreciative mind toward subjective knowledge
6. On objectivity, subjectivity, and probability
7. A delicate balance between order and chaos
8. The riddle of the ubiquitous statistics
Epilogue: Toward a new perspective on statistical inference

From the Intermission: "In the quest to develop "social physics" and "natural laws of society" researchers in the soft sciences have found that calculus is not readily applicable. In addition, nowhere can a grand unification like Newton's theory be anticipated. Researchers thus turn to a suspect area of mathematics that is considered as a "social calculus" (or "psychology calculus") but is more widely known as statistics."
At some point more than half-way through a specialist masters course in statistics I would make copies of this book available to the students. I would then set each of the various chapters as an essay topic for one student and towards the end of the course arrange a discussion in which the students could comment on the mistakes which the author accuses others of making, and also on the mistakes the author makes him-self. Such an exercise should entertain those taking part, and serve as a check to the over-emphasis on mathematics which is an almost inevitable feature of any such course.
The range of topics touched on, from the puzzles of quantum theory to statistical forecasting, and from discussions of causality to the various approaches to Total Quality Control, is a remarkable feature of the book.

Contents:
1. Preliminary information
2. Families of discrete distributions
3. Binomial distribution
4. Poisson distribution
5. Negative binomial distribution
6. Hypergeometric distributions
7. Logarithmic distribution
8. Mixture distributions
9. Generalized (stopped-sum) distributions
10. Matching, occupancy, and runs distributions
11. Miscellaneous discrete distributions

This is a revision of what has become a classic reference text. The expansion from 328 pages to 564 pages (an addition of 70% of the original text) is indicative of the amount of new material that has been included. There is also the addition of a third author from the first edition.
The Preface provides a guide as to what changes and extensions have been made. However, the twenty-year period from the appearance of the first edition to the present one may provide the best motivation for a revision. Much has happened during this period.
Chapter 1 provides preliminaries - mathematical, probabilistic, statistical, computational. This will be handy for various definitions of orthogonal polynomials, hypergeometric series, or generating functions, for example. Chapter 2 deals with families of discrete distributions. These families arise from an attempt to build an umbrella family that encompasses many special cases. For example, difference equations of the form Px+1=g(x)Px, where Px=P(X=x), yield a variety of special cases depending on the choice of g(x).
The binomial and Poisson distributions are the subjects of Chapters 3 and 4. These are the workhorses of discrete distributions and a number of interesting additions have been made. For example, a discussion of misrecorded Poisson distributions is included. The geometric distribution is discussed as a special case of the negative binomial distribution (Chapter 5). Because of its central role, I would have preferred a separate chapter on the geometric distribution, but this is a matter of taste in how to deal with a special case.
Chapters 5, 6 and 7 cover the hypergeometric distribution and the logarithmic distribution. Mixtures, the subject of Chapter 8, have proven to be extremely useful, and help provide an understanding of how some distributions arise. They also serve as a basis for generalizing to higher dimensions.
Chapter 9 on generalized stopped-sum distributions is new and should prove useful for some applications. Matching, occupancy, and runs distributions are the subjects of Chapter 10. This, too, is new to a great degree. Many occupancy models are based on the multinomial distribution. Since the authors have confined themselves to the univariate case, these more extensive models are not discussed. The final chapter provides a discussion of several special distributions,
such as the zeta distribution.
The index has been extended so that it is relatively easy to find a subject. The authors have taken a daring and precarious step by attributing many of the distributions to a particular individual. Where-as one might not wish to argue about the "Lagrangian binomial distribution", or the "super-Poisson distribution," there may be controversy over a "Hassenpfeffer distribution." Did Hassenpfeffer really discover this distribution, or was the germ of the idea in a previous paper by someone else?
A reviewer is supposed to point out failings as well as merits. The merits will be obvious to any reader who needs to use this text, and I did not find any failings. Indeed, the book is a gem. This book is indeed a magnum opus, and will, almost certainly, remain as the definitive reference for years to come. The volume is so extensive that the authors need not be concerned with the task of preparing for a third edition. They are to be thanked for their efforts, and congratulated for a job well done.

Contents:
1. Mathematical preliminaries
2. The G-function
3. Elementary special functions and the G-function
4. Generalizations to matrix variables

Readership: Graduate students and researchers in statistics and in the physical sciences

This book on special functions is accessible to the tenderfoot with a basic knowledge of mathematics as well as to the troop leader familiar with complex analysis. Chapter 1 gives the fundamental properties of the gamma, beta, psi and zeta functions; computational aspects are stressed, but rigorous proofs are regarded as freely available and are omitted. The emphasis is on continuous statistical density functions; a twelve page list is followed by ten pages of general theory. Chapters 2 and 3 treat Meijer's G-function and its special cases. These include generalized hypergeometric functions and hence the incomplete gamma, incomplete beta and error functions, and Chebyshev and Her-mite polynomials, etc. Here the reader is given much more help with the mathematics. In Chapter 4 the gener-al theory in Chapters 2 and 3 for scalar arguments is extended to matrix arguments; the matrix variate gamma, beta and Dirichlet densities are studied. Each chapter ends with twenty or more exercises.
The short bibliography does not try to cover the large body of research papers and algorithms on applications of special functions, but there are many applications in the text.
Comparison with the Handbook of Mathematical Functions, by M. Abramowitz and I.A. Stegun and Higher Transcendental Functions by H. Bateman and A. Erdélyi indicates that the coverage is not comprehensive; for example, it is hard to find mention of Euler polynomials or q-series. The strength of the book is its dedication to assisting the non-specialist to under-stand the potentialities of special functions and to helping the reader with their manipulation and computation.

Contents:
1. Statistical process control: A brief overview
2. Acceptance-rejection SPC
3. The development of mean and standard deviation control charts
4. Sequential approaches
5. Exploratory techniques for preliminary analysis
6. Optimization approaches
7. Multivariate approaches

Readership: Intended for "persons at all levels of mathematical ability." However this reviewer thinks the book is better suited to university courses than to brief training sessions in industry

The preface to this book is the most intriguing of any I have ever encountered in a statistical process control text. The authors use the back-drop of post-communist Eastern Europe to argue passionately for sweeping change in industrial management, especially the adoption of the fundamental principles of quality improvement of Pareto, Shewhart, and Deming. The book grew out of a series of short courses given to workers, foremen, and managers in companies in the United States of America and Poland. To those course notes, stronger theoretical underpinnings were added for each topic, and appendices on linear models and mathematical statistics were also included to make the book self-contained. Consequently the book is now more suited to advanced undergraduate and graduate university students of statistics than to industrial practitioners, who were the authors' original customers. The selection of topics is also much broader than customary for a statistical process control text, with traditional topics occupying barely half of the book. For example, Chapter 6 on optimization approaches covers simplex optimization, linear models, least squares, factorial designs, rotatable quadratic designs, saturated designs and simulation. The book is intellectually stimulating, sometimes on non-statistical issues, which is a rare bonus for readers in this field.

Contents:
1. Introduction
PART I : Basic Sampling
2. Simple random sampling
3. Confidence intervals
4. Sample size
5. Estimating proportions, ratios and subpopulation means
6. Unequal probability sampling
PART II : Making the Best Use of Survey Data
7. Auxiliary data and ratio estimation
8. Regression estimation
9. The sufficient statistic in sampling
10. Design and model
PART III: Some Useful Designs
11. Stratified sampling
12. Cluster and systematic sampling
13. Multistage designs
14. Double sampling
15. Network sampling
PART IV : Detectability Methods for Elusive Populations
16. Detectability and sampling
17. Line transects and variable circular plots
18. Capture-recapture sampling
19. Line-intercept sampling
PART V : Spatial Sampling
20. Spatial prediction or kriging
21. Spatial designs
22. Plot shapes and observational methods
PART VI : Adaptive Sampling
23. Adaptive sampling designs
24. Adaptive cluster sampling
25. Systematic and strip adaptive cluster sampling
26. Stratified adaptive cluster sampling

Readership: Statisticians, survey research workers and any scientific researcher who obtains data through sampling methods, including environmental scientists, geographers, geologists and archaeologists

This book is in six parts, the first three of which could be used as a course in sampling methods at the undergraduate level. The remaining three parts cover more specialized topics suitable for those with research interests in sampling from rare populations and spatial sampling. The book is a combination of a course text and a manual of sampling methods. Each topic is adequately explained with examples. For instance it would not be difficult for a geographer, to gain insight into the ideas of line transect sampling by looking at Chapter 17 without the need for frequent reference to other sections. This book is a joy to read; the explanations are clear and concise, the presentation is extremely pleasing, the references are up to date and there is an abundance of examples both in the text and at the end of each chapter.

Contents:
1. Introduction
2. Sources of survey error
3. Frames: Definitions of frames and frame errors
4. Frames: Quantifying frame errors
5. Frames: Conducting surveys with imperfect frames
6. Nonresponse: Background and terminology
7. Nonresponse: Statistical effects of the problem
8. Nonresponse: Dealing with the problem
9. Measurement: Survey measurement and measurement error
10. Measurement: Quantifying measurement error
11. Measurement: Quantifying measurement error, variability in measurement
12. Total survey design: More general error models

Readership: Researchers and graduate students in sample survey theory, survey practitioners

This book gives an excellent summary of the huge literature on nonsampling errors in surveys. Sources of error are grouped under three general headings: frame errors, nonresponse errors and measurement errors. Under each heading, the main concepts are defined, likely effects are reviewed and suggested remedies are explored. It is not intended to be a detailed handbook on how to carry out a survey, but a broad overview of what can go wrong in a survey, and where. The authors have done an admirable job in collecting results from a wide variety of sources and putting them together in a single coherent framework with a standardized notation. It should become a standard reference in the area.

Contents:
1. Estimation in finite populations: A unified theory
2. A few sampling strategies
3. Predictors
4. Asymptotic approaches
5. Design- and model-based variance estimation
6. Multi-stage, multi-phase and repetitive sampling
7. Re-sampling and variance estimation in complex surveys
8. Sampling from inadequate frames
9. Analytic studies of survey data
10. Randomized response
11. Incomplete data
12. An epilogue

Readership: Researchers and graduate students in statistics, survey practitioners with a strong interest in the underlying theory
This book gives a concise account of modern sample survey theory, with good coverage of the developments during the last two decades. The authors have done an admirable job in collecting results from a wide variety of sources and putting them together. The cover suggests that the book will be a valuable reference for sociologists, demographers, biostatisticians, economists and public administrators, but most practitioners may find the theoretical level rather forbidding.

Contents:
1. Introduction
2. A general linear mixed model
3. First order autoregressive models
4. State space representations
5. The Laird-Ware model in state space form
6. Autoregressive moving average errors
7. Nonlinear models
8. Multivariate models

Readership: Students at the graduate level in biostatistics, statistics or other disciplines that collect longitudinal data

The last ten years or so have witnessed great strides in the development of techniques for analyzing longitudinal data. These have been summarized in four books published in the last couple of years, including Analysis of Repeated Measures, by M.J. Crowder and D.J. Hand, 1990, [Short Book Reviews, Vol. 10, p.45.]. The present book is written in a very readable style and the mathematics is not needlessly complicated. It
focusses on models with Gaussian error structures and adopts the maximum likelihood approach. The first three chapters provide the background and Chapters 4 to 8 describe approaches using the state-space representation and the Kalman filter technique for computing likelihoods with correlated data. The book has the merit of being short, but this is at the cost of there being relatively few real-data examples. There are short, theory-based exercises at the end of all but Chapter 7.
I enjoyed this book and recommend it, especially in combination with one of the other books which provide real-data examples.

Contents:
1. Preview
2. The sample survey
3. Other sampling designs
4. The linear regression model
5. Experimental designs to assess the effect of a treatment
6. Interrupted time series
7. More advanced experimental designs
8. Some special types of data
9. Computer-intensive statistics
10. Ethical considerations
11. Synthesis: Carrying out a research study

The design and the analysis are both essential aspects of research studies. This is by now essentially universally accepted both by statisticians and by re-search workers with a background in the substantive areas of application, and yet disagreements arise and defective studies are published. Why should this be? Perhaps because of difficulties in understanding variation, and a reluctance to admit how complicated cause-and-effect relationships can be. The present book deals, sensibly but inevitably rather briefly, with a wide variety of types of situation, for example randomized experiment and observational study, and of methods of collecting data, for example surveys, capture-recapture methods, classical designs, illustrating many of these and the points they raise by examples and case studies.

Contents:
Preface: A brief history of the movement to study cognitive aspects of surveys and the Social Science Research Council (SSRC) Committee
PART I : Introduction
1. Cognitive aspects of surveys and this volume
PART II : Meaning
2. Asking questions and influencing answers
3. Direct questioning about comprehension in a survey setting
PART III: Memory
4. Personal recall and the limits of retrospective questions in surveys
5. Improving episodic memory performance of survey respondents
6. Memory and mismemory for health events
7. Attempts to improve the accuracy of self-reports of voting
8. Applying cognitive theory in public health investigations: Enhancing food recall with the cognitive interview
PART IV : Expression: The Case of Attitude Measurement in Surveys
9. Opportunities in survey measurement of attitudes
10. The case for measuring attitude strength in surveys
11. New technologies for the direct and indirect assessment of attitudes
PART V : Social Interaction
12. Validity and the collaborative construction of meaning in face-to-face surveys
PART VI : Government Applications
13. A reveiw of research at the Bureau of Labor Statistics

This book is a collection of papers that focuses on the contributions of the cognitive sciences to understanding issues in survey research. It brings together some of the theory and empirical research that have developed out of the movement to study cognitive aspects of surveys. The book is the product of the Social Science Research Council Committee on Cognitive and Survey Research. The material for most of the chapters originated in the committee's meetings and workshops. Particularly interesting are Part II on "Meaning" that reflects the view that respondents must share the meaning of a question intended by the survey researcher if they are to respond usefully and Part III on "Memory" that deals with respondents' retrieval of information. The book will be of interest to survey researchers and questionnaire designers who want to learn more about the cognitive bases of survey responding.

Contents:
Introduction to Statistics in the Environmental Sciences
1. Establishing a statistical context for the evaluation of air quality models, A.D. Thrall
2. Quality data networks that minimize entropy, W.F. Caselton, L. Kan and J.V. Zidek
3. Nonparametric estimation of spatial covariance with application to monitoring network evaluation, P. Guttorp, P.D. Sampson and K. Newman
4. Spatial covariance estimation for monitoring data, C. Louder and P. Switzer
5. Modeling daily precipitation - progress and problems, D.A. Woolhiser
6. Aging functions and their nonparametric estimation in point process models of rainfall, M.J. Phelan Introduction to Statistics in the Earth Sciences
7. Deconvolution of chaotic and random time series, J.D. Sargle
8. Deconvolving nonGaussian time series: The seismic experience, A.T. Walden
9. Deconvolution of short period teleseismic and regional time series, Z.A. Der, A.C. Lees, K.L. McLaughlin and R.H. Shumway
10. Envelope estimation for quasi-periodic geophysical signals in noise: A multitaper approach, J. Park
11. Statistical methods for the description and display of earthquake catalogues, D. Vere-Jones
12. Stochastic models for the distribution of rock types in petroleum reservoirs, B.D. Ripley
13. Analysis of geochemical data sampled on a regional scale, K. Conradson, A.A. Nielsen and K. Windfield

This is a compendium of papers which the editors solicited from the participants in a summer 1989 conference which was held in Belgium and entitled "Statistics: Space and Earth Sciences". Papers in this volume are organized in two groups corresponding to the physical domain of the problems discussed. The lead paper raises a challenge to include the effects of model uncertainty in inferences which are made with the output from geophysical models. The author points out that policy makers, and other model users, need to be able to quantify the uncertainties which arise from a less than complete understanding of the physical processes represented by the model and from the often numerous parameterizations of unresolved processes which are part and parcel of models of geophysical systems. This challenge will only be answered through unprecedented interaction between geophysical scientists and statisticians. The remainder of the volume is more technical in nature and is illustrative of some of the best current work in the statistical sciences to emerge from geophysicist/statistician interactions. Statisticians will find these papers to be both accessible as there is a minimum of subject matter jargon, and substantive. Unfortunately, for the same reasons, many geophysical scientists will find these papers to be less engaging.

Contents:
1. An introduction to classification and clustering
2. The initial examination of multivariate data
3. Measurement of similarity, dissimilarity and distance
4. Hierarchical clustering techniques
5. Optimization methods for cluster analysis
6. Mixture models for cluster analysis
7. Other clustering techniques
8. Some final comments and guidelines

This is a brief, but lucid and very readable, introduction to the methodology of cluster analysis. It describes most of the important approaches and techniques, as well as briefly discussing multivariate plotting and ordination, and touching upon aspects such as measurement and comparison of classifications. Mathematical formulae are kept to a minimum, and the emphasis is towards practice rather than methodology. Interesting applications, and clear step-by-step discussion of simple examples, are among the book's strengths. Much new work has appeared over the thirteen years since the previous edition, and sixty-three of the two hundred and thirty-two references in the new edition are to papers published after 1980. Perhaps surprisingly, none of the new references comes from the Journal of Classification (established in 1984 for developments in this general area). Also, the reader is referred "elsewhere for details" rather often. However, these are minor quibbles. Overall, the text provides an excellent introduction to an important topic.

Contents:
PART I : Fundamentals
1. Introduction
2. Some key ideas
3. Experimental strategies
4. The choice of a model
5. Models and least squares
6. Criteria for a good experiment
7. Standard designs
8. The analysis of experiments
PART II : Theory and Applications
9. Optimum design theory
10. Criteria of Optimality
11. D-optimum designs
12. Mixture experiments
13. Experiments with both qualitative and quantitative factors
14. Blocking response surface designs
15. Algorithms for the construction of exact D-optimum designs
16. Restricted design regions
17. Failure of the experiment and design augmentation
18. Non-linear models
19. Optimum Bayesian designs
20. Discrimination between models
21. Composite design criteria
22. Further topics

The first part of this book provides a condensed but authoritative statistical overview of the practice and theory of modelling, designing and analyzing systematic experiments. Where there is selectivity it is directed towards the requirements of the second part, which forms the bulk of the book and is principally concerned with D-optimality linked to the General Equivalence Theorem. The techniques, many of which were developed by the authors, are clearly described and demonstrated in examples which are mainly of an industrial type. There are no exercises for the reader. There is a relatively short but useful Fortran program for constructing D-optimal designs listed in an appendix.

Contents:
1. Introduction
2. Observational errors
3. Linear least squares
4. Nonlinear least squares
5. Formulation and selection of models
6. Criteria for model selection
7. Polynomials
8. Fitting functions
9. Fourier transform techniques
10. Potentiometric titrations

Readership: Those interested in non-linear model fitting, particularly in chemical science applications

The author has written an attractive short book which efficiently describes both linear and non-linear least squares plus some specific applications in which he has had experience extending over more than twenty years. Matrix algebra is used throughout where relevant; and the discussion is brisk and authoritative. This might be a good place for chemical science readers to pick up the material, and it could also provide supplementary reading to recent, more complete statistics textbooks on nonlinear estimation by, for example, D.M. Bates and D.G. Watts Nonlinear Regression Analysis and its Applications [Short Book Reviews, Vol.9, p.5], A.R. Gallant Nonlinear Statistical Models [Short Book Reviews, Vol.7, p.29] and G.C. Wild and G.A.F. Seber Nonlinear Regression [Short Book Reviews, Vol.9, p.48]. It is not a competitor of these. There are no exercises and it is a monograph for seminars
rather than a class text, unless supplemented.

Contents:
1. Introduction
2. Direct assays
3. The indirect quantitative assays
4. Quantal assays: Single agent
5. Other methods for estimating LC50
6. Quantal assays: Comparative studies
7. Multivariate approach to bioassays

Readership: Students and research workers in the biological sciences and biostatisticians

This book is intended as a text for a course in bioassay methods. It has good numerical examples for most of the techniques presented. There is also a set of problems included at the end of each chapter to reinforce the concepts presented. The material is presented with reasonable mathematical rigour with many of the proofs presented in appendices.
The third edition includes an updated bibliography with 929 references related to bioassay. In order to make this extensive bibliography more useful, an excellent index is included. This edition also adds a discussion of logit analysis and drops a section on computing methods.

Contents:
1. Introduction
2. Some basic considerations concerning estimation in clinical trials
3. The AB/BA design with normal data
4. Other outcomes and the AB/BA design
5. Normal data from designs with three or more treatments
6. Other outcomes from designs with three or more treatments
7. Some special designs
8. Graphical and tabular presentation of cross-over trials
9. Various design issues
10. Mathematical approaches to carry-over

This practical book is a good introduction to the controversial topic of cross-over trials in pharmacological studies for both medical scientists and statisticians. Methods of analysis are explained using heuristic rather than mathematical arguments, and illustrated through many examples from the author's own experience in the pharmaceutical industry. To aid understanding, some analyses are presented in terms of hand-calculator methods before SAS computer analysis is outlined. The author is careful to distinguish between standard approaches, presented in a lively, practical way, and his personal views. For example he argues that there is no place, in the design and analysis of pharmacological cross-over studies, for the widely-used linear model in which carry-over from
preceding treatments is described by an additive first-order carry-over effect. These views make the book a thought-provoking read for all statisticians interested in cross-over studies.

Contents:
1. Introduction to the analysis of cross-over designs: Basic principles and some useful tools
2. Latin square designs
3. The 2-treatment, 2-period, 2-sequence design
4. Modification of the 2-period, 2-treatment, 2-sequence design
5. Cross-over designs with variance balance
6. Cross-over designs lacking variance balance
7. The analysis of categorical data from cross-over designs
8. Ordinary least squares estimation versus other criteria of estimation
9. Other topics in cross-over designs

Readership: Statisticians/researchers in psychology, agriculture, social science, medicine

The authors take an approach which is completely different from that of the previous book [Short Book Reviews, Vol.13, p.21] in that they consider experiments for which a simple first-order carry-over model is a reasonable description of the data. Much of the book is devoted to showing how SAS can be used to fit such models with an emphasis on testing parameters, rather than estimating the contrasts of interest. Some of the examples presented would benefit from the use of graphical data displays and more discussion of the practical interpretation of the results. Two chapters, which practitioners may find particularly useful, catalogue a large number of designs from the literature together with a variety of assessments of their performance in estimating direct effects under the assumed model. There are some confusions and inaccuracies, for example balance and orthogonality are not clearly de-fined and their relationship is wrongly stated.

Contents:
1. Introduction
PART I : Probability
2. Concepts of probability
3. Theory of probability
PART II : Statistics
4. Concepts of statistics
5. Theory of statistics: Introduction
6. Estimation and localization
7. Hypothesis testing
PART III: Decisions
8. Decision theory
9. Utility theory
10. Multiple decisions
11. Sequential methods

The purpose here is to provide sufficient probability and statistical theory for a reader with basic calculus to understand the decision theory topics. The treatment is lack-lustre and tends to be a little idiosyncratic. What is the point of including the Borel-Cantelli lemma at this level? Whilst explanations are generally acceptable, on a number of occasions they degenerate, as for instance when the Central Limit Theorem is introduced in the form "the average value of any collection of random variables converges to a particular density function". Given that the intended audience consists, in part at least, of non-statisticians, the material is not well motivated. Some case studies in the decision theory part might have helped; there does not appear to be a single genuine set of data in the entire book. Although it might be argued that this is, after all, an introduction to decision theory and not applied decision theory, an objective of the author was to appeal to those non-statisticians "who desire to apply decision-theoretic thinking to their own work". Such readers would end up with little idea as to the usefulness or limitations of the methods in practice.

Contents:
1. Probability spaces and random variables
2. Independent identical trials and the law of large numbers
3. De Moivre-Laplace and Poisson limit theorems
4. Conditional probability and independence
5. Markov chains
6. Random walks on the lattice Æd
7. Branching processes
8. Conditional probabilities and expectations
9. Multivariate normal distributions
10. The problem of percolation mathematical expectation
11. Distribution functions, Lebesgue integral and
12. General definition of independent random variables and laws of large numbers
13. Weak convergence of probability measures on the line and Helly's theorems
14. Characteristic functions
15. Central limit theorem for sums of independent random variables
16. Probabilities of large deviations

The text is organized in lectures and contains some interesting topics, not usually included in a book for beginners: for example, percolation, stability in the central limit theorem, Monte Carlo methods. The last six lectures reach a higher level than the first part of the text. The lack of exercises is a short-coming, and makes the book less suited for self-study.

Contents:
1. Introduction
2. Resampling-based adjustments: Basic concepts
3. Continuous data applications: Univariate analysis
4. Continuous data applications: Multivariate analysis
5. Binary data applications
6. Further topics
7. Practical applications

Readership: Research workers, practising and research statisticians, graduate level students

In many contexts there are good reasons, as well as bad, for performing multiple hypothesis tests. But there are pitfalls in doing so, and these are all the greater since the suite of statistical tools avail-able for controlling multiplicity, through adjustment of p-values for the particular set of questions asked, is somewhat limited. This clearly written and nicely presented book advocates a general solution to the multiplicity dilemma through the use of resampling methods. In a treatment both accessible to practitioners yet of interest to bootstrap specialists, the authors demonstrate how resampling methods can be developed to analyze a wide range of models and data structures. Though the development is detailed and scholarly, the emphasis is on implementation. The book contains many examples and applications. Many of the analyses are carried out using the SAS MULTTEST procedure, which is detailed in an appendix. The book illustrates implementation of the algorithms in other high-level programming languages. This is a good book. Success will be judged by the extent to which practitioners adopt the procedures described. This they ought to do.

Contents:
0. Introduction
1. Bootstrap and asymptotic normality
2. An example where bootstrap fails: Comparing nonparametric versus parametric regression fits
3. A bootstrap success story: Using nonparametric density estimates in k-sample problems
4. A bootstrap test on the number of modes of a density
5. Higher-order accuracy of bootstrap for smooth functionals
6. Bootstrapping linear models
7. Bootstrapping robust regression
8. Bootstrap and wild bootstrap for high-dimensional linear random design models

Of key interest to statistical practitioners who may wish to use the bootstrap is knowledge of circumstances where it will provide valid inference. This volume of lecture notes goes some way towards a clear setting down of the conditions under which one can expect the bootstrap to work satisfactorily. Based closely on the published papers of a leading researcher in this important area, it is concerned primarily with proofs of the asymptotic behaviour of the bootstrap in a variety of contexts. Theoretical results on application of the bootstrap to linear models, nonparametric curve estimation and estimation of smooth functionals are supplemented by substantial simulation studies. It is notable that these often provide warnings against over-reliance on the asymptotic theory. The specialist will find this a useful drawing together of research papers. The non-specialist will find the treatment very specific.

Contents:
PART I : Information Bounds
1. Models, scores, and tangent spaces
2. Convolution and asymptotic minimax theorems
3. Van der Vaart's differentiability theorem
PART II : Nonparametric Maximum Likelihood Estimation
1. The interval censoring problem
2. The deconvolution problem
3. Algorithms
4. Consistency
5. Distribution theory

Readership: Research workers in non-parametric and semi-parametric statistical theory

This volume of lecture notes consists of two parts, one written by each author. Part I, based on Wellner's lectures, gives a brief sketch of information lower bound theory in non-parametric and semi-parametric models, including Hajek's convolution theorem and its extensions, and a result characterizing differentiable statistical functionals by van der Vaart. The level of abstraction is quite high, but the results are well presented and important to this area. Part II, based on Groeneboom's lectures, focuses on the more practical side of nonparametric maximum likelihood estimation, with emphasis on certain censoring and deconvolution problems. The developments include characterization of the solutions using ideas from isotonic regression, together with the corresponding algorithms for their computation, and some asymptotic consistency and distribution theory. Both authors write in a clear and organized fashion, and the material presented covers many important aspects of this challenging, but important, subject.

Contents:
1. Introduction
2. Definitions and basic properties of Voronoi diagrams
3. Generalizations of the Voronoi diagram
4. Algorithms for computing Voronoi diagrams
5. Poisson Voronoi diagrams
6. Spatial interpolation
7. Models of spatial processes
8. Point pattern analysis
9. Locational optimization through Voronoi designs

Readership: Advanced students and researchers in spatial processes, computational geometry and related fields

This book is most welcome and very comprehensive, a reference book in general but also a good read for those interested in particular chapters. The emphasis is on statistical matters: point processes, multivariate rather than pure mathematics, packing, covering, etc. This reviewer would have liked more on the higher-dimensional computational geometry, but everything included is excellent. There are beautiful two-dimensional diagrams which are "worth a thousand words".

Contents:
1. Prolegomena
2. The spatial random field model
3. The intrinsic spatial random field model
4. The factorable random field model
5. The spatiotemporal random field model
6. Space transformations of random fields
7. Random field modeling of natural processes
8. Simulation of natural processes
9. Estimation in space and time
10. Sampling design

The book treats random field models with a view to applications in the earth sciences and other physical sciences. Therefore, the emphasis is on the type of models used in the so-called geostatistics and on spatiotemporal processes. Particular attention is given to relations to stochastic differential equations. Empirical validation, prediction and simulation of the models as well as sampling design are treated, whereas other questions of statistical inference, for example, parameter estimation, are hardly discussed. Point processes and Markov random fields are not con-sidered. The book is an interesting supplement to other books on random field models, but unfortunately the mathematics is not always well presented, and life is made unnecessarily difficult for the reader by the excessive use of abbreviations. The reader is expected to know about things like Hilbert spaces, conjugate operators, Schwartz spaces and generalized functions. There are several philosophical sections which do not go deeply into the problems, but do contain some statistical common sense. I particularly like the strong emphasis on the mathematical model as a principal component of empirical investigation.

Contents:
1. Introduction
2. The mathematic background
3. Model specification and censoring
4. Nonparametric estimation
5. Nonparametric hypothesis testing
6. Parametric models
7. Regression models
8. Asymptotic efficiency
9. Frailty models
10. Multivariate time scales

Readership: Statisticians, mathematically inclined actuaries, demographers, epidemiologists, etc.

The past two decades have produced enormous advances in the theory and methods of survival analysis and, more generally, event-history analysis. The two major influences on this development were D.R. Cox's 1972 and 1975 papers on life tables, regression and partial likelihood, and O.O. Aalen's 1975 thesis and 1976-78 papers casting event-history analysis in the framework of counting processes, martingales and stochastic integration. This book is a masterful account of the counting process approach to event-history analysis by four of its leading contributors. The emphasis is on non-parametric and semi-parametric methods (although there is a chapter on parametric models) and the book relies heavily on martingales and stochastic integration, so is by no means all easy reading. However, the authors have produced a superb blend of real examples, important practical discussion, methodology and theory. This comprehensive work of scholarship is certain to be the standard reference for the area, and should be on the bookshelf of anyone interested in event-history analysis.

Contents:
PART I : Foundations
1. Random variables
2. Probability measures
3. Stochastic processes
4. The stochastic integral
PART II : Existence and Uniqueness
5. Linear equations with additive noise
6. Linear equations with multiplicative noise
7. Existence and uniqueness for nonlinear equations
8. Martingale solutions
PART III: Properties of Solutions
9. Markov properties and Kolmogorov equations
10. Absolute continuity and Girsanov's theorem
11. Large time behaviour of solutions
12. Small noise asymptotics

From the book cover: "The aim of this book is to give a systematic and self-contained presentation of basic results on stochastic evolution equations in infinite dimensional, typically Hilbert and Banach, spaces." This describes the purpose of the book precisely. There was quite a need of such a fine and explicitly written book about this area of probability theory. It is appealing to both probabilists and analysts. An essential role in its approach to the subject is played by both semigroup theory and control theory. An appendix containing additional basic material, used in the main text, makes the book indeed reasonably self-contained. A rich bibliography provides some guidance for the interested reader through the enormous literature in the field. This book is recommended for the probability section of every mathematical library.

Contents:
1. Strong approximations of partial sums of independent identically distributed random variables
2. Renewal and related processes
3. Uniform empirical and quantile processes and their approximations
4. Weighted approximations of uniform empirical and quantile processes
5. Asymptotic distributions of functionals of weighted uniform empirical and quantile processes
6. General quantile processes and their approximations

A very important methodology for obtaining strong and weak approximations for various stochastic processes was developed around 1975 in the fundamental papers by M. Csörgö and P. Révész and also by J. Komlós, P. Major and G. Tusnády, which is now called the 'Hungarian construction'. The first book on this topic was written by M. Csörgö and P. Révész in 1981 Strong Approximations in Probability and Statistics, [Short Book Reviews, Vol.2, p.32] and the present monograph by M. Csörgö and L. Horváth includes an update of some major advances in the past decade. Chapter 1 provides a very useful survey and discussion of the Komlós-Major-Tusnády theorems for partial sums of independent and identically distributed random variables and the optimality of the resulting almost sure approximation rates. Chapter 2 deals with approximations for the renewal process and related processes and gives brief applications to, for example, queueing and risk theory. Chapter 3 considers two other processes that are each other's inverse: the uniform empirical and quantile processes. Weighted approximations come into the picture at about midway in the book. In Chapter 4 the authors consider best possible weighted approximations for the uniform empirical and quantile processes. Asymptotic distributions of functionals of these are treated in Chapter 5. The final chapter, Chapter 6, gives the theory for samples for a general distribution instead of the uniform. The style of the monograph is, unavoidably, technical and dry. Fortunately, there are enough parts with useful remarks, intuitively explained ideas, historical notes, etc. which make the reading more pleasant. There is no doubt that researchers in the field will be very happy with this book and its useful list of more than two hundred bibliographical references.

Contents:
1. Univariate random variables
2. Multivariate random variables
3. Limit laws
4. Markov chains: Passage phenomena
5. Markov chains: Stationary distributions and steady state
6. Markov jump processes
7. Ergodic theory with an application to fractals

A special feature of this book is the inclusion of twelve colour graphs of fractals. Their presentation is motivated by a discussion of products of random matrices in Chapter 7. The first six chapters rely heavily on existing material in standard text-books, all being properly referenced and acknowledged. An interesting feature is the detailed discussion of Oseledec's theorem on the spectral decomposition of products of random matrices which, as far as I know, appears for the first time in an introductory book on stochastic processes. Although I am left with the impression that the book has been written around the pictures, I am sure that Chapter 7 will be useful to complement existing courses on the subject.

Contents:
1. Stopping times
2. Infinite measure and Orlicz spaces
3. Inequalities
4. Directed index set
5. Banach-valued random variables
6. Martingales
7. Derivation
8. Pointwise ergodic theorems
9. Multiparameter processes

Readership: Students and researchers in probability theory, analysis and ergodic theory

If you read only the first half of the title "Stopping times", then you will guess that this is a book about martingales. If you read the second half "directed processes", then you will think that it is very abstract processes with very abstract parameter spaces. If you do not like very abstract things perhaps you say that it is enough. But believe me, you make a big mistake.
If you read further you will realize that it is a very well-written book, explaining why the treated questions are interesting. Even the technical details of the proofs are presented in a very intelligible form.
The authors show how their abstract results can be applied to obtain very concrete results. Beautiful examples include the surprising prophet in-equality, the large numbers for star-mixing processes and the unification of martingale and ergodic theorems.

Contents:
1. Markov processes
2. Examples
3. Point processes of excursions
4. Brownian excursions
5. Itô's synthesis theorem
6. Excursions and local time
7. Excursions away from a set

The basic idea in excursion theory centres around a path-decomposition of general stochastic processes which, when properly interpreted, yields an imbedded structure akin to that of Lévy processes. This procedure is worked out in full detail for Markov processes. The book is to a large extent self-contained. Many interesting examples help the reader in getting a deeper understanding of these fundamental results from the modern theory of stochastic processes. New theory is always well-motivated and more advanced theoretical discussions are compensated by very helpful general comments. The result is an excellent and scholarly-written text which many of us will find fascinating to read and ideal to teach from.

Contents:
1. Discrete-event simulations
2. Regenerative stochastic processes
3. Regenerative simulation
4. Networks of queues
5. Passage times
6. Simulations with simultaneous events
APPENDIX A: Limit theorems for stochastic processes
APPENDIX B: Random number generation

Based upon the theory of generalized semi-Markov processes, a simulation methodology using imbed-ded regenerative structures of such processes is presented. Besides yielding the generation of relevant sample paths, the method allows for statistical tests on the simulation output. The author strikes an ideal balance between theory and practice. The necessary theoretical discussions are compensated by numerous examples from diverse fields of engineering. Various algorithms which help computer implementation are given. Each chapter contains taxing, but at the same time very interesting exercises, each of which contains a non-trivial simulation experiment. Two extensive appendices make this text essentially self-contained. The result is a well-structured, very readable and informative text which will be useful to a wide audience of students and researchers interested in simulation methodology.

Contents:
1. Stochastic models for random sets of points
2. Poisson processes in general spaces
3. Sums over Poisson processes
4. Poisson processes on the line
5. Marked Poisson processes
6. Cox processes
7. Stochastic geometry
8. Completely random measures
9. The Poisson-Dirichlet distribution

As the author notes, Poisson processes are fundamental in both theoretical probability and proba-bilistic modelling. Their general structure is simple and elegant but much less widely known than it should be. First acquaintance with Poisson processes is usually in one dimension and in that setting (perhaps as with Lebesgue integration) special properties of the real line obscure the general picture. This book will do much to redress the balance. It provides an enjoy-able and clearly written introduction to the structure and properties of Poisson processes. Thanks to skillful steering away from, and around, technicalities, it is widely accessible. If you don't know the story, read this book - you will then know what you are missing. If you do know it, a browse through the book, particularly the later chapters, is still worthwhile for interesting perspectives on several areas.

Contents:
1. Modèles stationnaires au second ordre sur Zd
2. Champs de Gibbs et champs Markoviens
3. Théorèmes limites et estimation paramétrique pour les champs
4. Estimation pour les processes au second ordre
5. Estimation des champs de Gibbs
6. Algorithmes stochastiques

Markov and other multidimensional random fields, and the statistical theory for them, have be-come very important in applications to spatial problems, like image reconstruction. The main aim of this book is to give a mathematically sound introduction into the field, presenting also all the probabilistic background on spectral theory, Gibbs fields, central limit theorems and other topics. Next, the author gives an overview of the most important statistical methods of, for example, pseudo-likelihood estimation. An important role is played by Monte Carlo methods like simulated annealing and other stochastic algorithms. The book closes by going into some concrete applications to image reconstruction. Perhaps, the part with the applications is a bit too short, but otherwise this is a very nice introduction into the field.

Contents:
Introduction
PART I : Statistical Expert Systems
PART II : Belief Networks
PART III: Learning
PART IV : Neural Networks
PART V : Text Manipulation
PART VI : Other Areas

This book covers state-of-the-art work being carried out at the broadening frontier between statistics and artificial intelligence. It contains selected refereed papers from the Third International Workshop on Artificial Intelligence and Statistics, 1991. The first part presents work concerned with the development and use of expert systems to help re-searchers do statistical analysis and includes papers on experimental design and explanation. The remaining parts, in contrast, concentrate on ways in which statistical ideas and techniques have contributed to developments in various areas of artificial intelligence. These areas include belief networks, learning, knowledge acquisition, expert systems, non-monotonic reasoning, probabilistic logic, neural networks and text manipulation and understanding. The book provides a broad update of current work that should aid re-searchers on both sides of the AI/statistics interface, and is a useful successor to earlier volumes in the series from W.A. Gale (Ed.) (1986) Artificial Intelligence and Statistics, [Short Book Reviews, Vol.7, p.9.] and D.J. Hand (Ed.) (1990) Artificial Intelligence and Statistics, Volume II.

Contents:
1. Introduction
2. Random numbers
3. Sampling from non-uniform distributions
4. Economic and corporate models
5. Operations research models
6. Simulation software
7. Statistical applications
8. Regression metamodels
9. Design of experiments
10. Tactical aspects
11. Verification and validation
12. Epilogue

Readership: Students of management science, information systems, and others who wish to apply simulation

[This book is an adaptation of a Dutch version published in 1988].
The author describes Chapters 8 and 9 as setting the book apart from other textbooks on simulation. The first of these chapters describes the use of multiple regression models and the input/output behaviour of simulation models. The second describes the use of design of experiments to set the levels of the factors to use in the simulations. As statisticians we spend a good portion of our time telling researchers that they should collect the data properly according to some efficient and reliable design, but all too often one sees simulation studies reported in the statistical literature where these very aspects have been ignored, where little consideration seems to have been given to the relevant range of the factors in the simulation and where no effort has been made to use a design which will produce the most informative results. I am therefore delighted to see the emphasis in these chapters. A supplementary disc with a collection of exercises in Pascal is also available.
I found this an accessible book, although some of the notation did not seem ideal, such as the under-score to indicate a random variable. I recommend it to anyone interested in a quick and broad introduction to the subject.

Contents:
Introduction
1. Semigroup of operators and interpolation
2. Variational theory of parabolic systems
3. Semigroup methods for systems with unbounded control and observational operators
4. Differential systems with delays

Readership: Mathematicians and control theorists with an interest in infinite, dimensional systems

This book, in two volumes, is concerned with the study of linear and quadratic optimal control of infinite-dimensional systems. Volume 1 outlines the background systems theory for such systems and Volume 2 considers the associated optimal control problems. Potential applications of this theory occur in many areas, including control of large space structures, control of plasma fusion and chemical process control in the presence of long time-delays. The authors of this book are well known for their extensive contributions to the control of infinite-dimensional systems. The book is an invaluable reference source for this topic.

Contents:
1. Introduction
2. A general framework for Markov decision processes
3. Algorithms
4. Linear programming formulations for Markov decision processes
5. Semi-Markov decision processes
6. Partial observable and adaptive Markov decision processes
7. Further aspects of Markov decision processes
8. Some Markov decision process problems, formulations and optimality equations

The author follows a step-by-step approach in developing Markov decision processes emphasizing the essential link between linear programming and the solution of the optimality equations. He does so by including more modern topics such as: structural policy analysis, multiple objectives, utility and Markov games. The level of the text lies in between that of most elementary and advanced treatments. By including partly solved exercises, detailed examples and illustrated algorithms, the book should prove to be a ready-made textbook for a course on Markov decision processes.

Contents:
1. Netform origins and uses: Why modeling and netforms are important
2. Fundamental models for pure networks
3. Additional pure network formulation techniques
4. Dynamic network models
5. Generalized networks
6. Netforms with discrete requirements

This book describes a notation, albeit a visual one, for formulating network models. A wide range of practical applications, many based on the authors' extensive experience, are discussed to demonstrate the diversity of problems that can be modelled as networks. The text requires no mathematical ability. The text refers to 'sophisticated network codes' but gives no suggestions as to where such codes are avail-able today; the reader is left with the impression that they run on IBM 370 and CDC 6600 machines. Only a well-informed reader would realize that these models can be run, even if slowly, with the linear programming code on a workstation. There is no mention as to how to interpret and use the dual values that are a component of the optimal solution. Surely modelling means not only initially formulating the model but also under-standing the results and, if necessary, using them to refine the model. This is a very disappointing book.

Contents:
1. Introduction: A new age of dynamics
2. How to identify chaotic vibrations
3. Models for chaos, maps and flows
4. Chaos in physical systems
5. Experimental methods in chaotic vibrations
6. Criteria for chaotic vibrations
7. Fractals and dynamical systems
8. Spatio-temporal chaos
APPENDIX A: Glossary of terms in chaotic and nonlinear vibrations
APPENDIX B: Numerical experiments in chaos
APPENDIX C: Chaotic toys
APPENDIX D: Books on nonlinear dynamics, chaos and fractals

This is a new, much revised and enhanced, edition of the author's previous book on Chaotic Vibrations (1987). The emphasis in the general selection and treatment of material is very much as before, but with more applications, new research material, problems and many more references. The book contains a wealth of physical examples from science and engineering which illustrate how, and under what circum-stances, chaotic behaviour can be recognized and what consequences this may have. In particular, these consequences involve the 'sensitivity to initial conditions' characteristic of chaos and the apparently paradoxical way in which a system which is completely deterministic can produce output which is effectively 'stochastic' in nature.
From the point of view of applied mathematicians the great range of examples presented is highly rewarding and the mathematical detail they might desire can be found elsewhere. This is certainly the style of a book which deserves to make converts to the cause among applied scientists and engineers, since the scope of applicability of the ideas is impressive and not obscured by the mathematical detail which they might find discouraging. Appendix C is novel in my experience!

Contents:
1. Elements of decision under uncertainty
2. Risk-bearing: to the optimum of the individual
3. Comparative statics of the risk-bearing optimum
4. Market equilibrium under uncertainty
5. Information and informational decisions
6. The economics of emergent public information
7. Research and invention
8. Informational asymmetry and contract design
9. Strategic uncertainty and equilibrium concepts
10. The economics of contests
11. Competition and hidden knowledge
12. Long-run relationships and the credibility of threats and promises

This book consists of two untitled parts. Part I deals with the analysis of uncertainty faced by a decision maker, and with the implications of uncertainty on economics, risk-bearing and market equilibrium. In Part II, dynamic states of knowledge are discussed in what could be termed "the economics of information". The book is written in a refreshing style, it is easy to read and it contains many examples and discussions. It is difficult to judge whether the "global" theory forms a suitable model for the many practical applications presented by the authors. How-ever, the resulting uncertainty and information models are remarkably coherent and versatile.

Contents:
1. The mathematics of populations: Demographics
2. Inheritance
3. A theory of epidemics
4. Biogeography
5. The heart and circulation
6. Gas exchange in the lungs
7. Control of cell volume and the electrical properties of cell membranes
8. The renal countercurrent mechanism
9. Muscle mechanics
10. Biological clocks and mechanisms of neural control

This is a book in two distinct halves: Chapters 1-4 give a brief overview of a range of fields concerned with populations while Chapters 5-10 deal with some mathematical models in physiology. The former are at an elementary level while the latter are more mathematically sophisticated, requiring a `feeling' for detailed models involving differential equations. The intended audience is pre-medical science under-graduates, but the book will also prove useful to undergraduates in applied mathematics.
There are obvious difficulties inherent in a short, introductory book with such an ambitious scope. Models are formulated and then discarded, without al-lowing a full appreciation of what a mathematical model can and cannot contribute to an understanding of the phenomenon concerned. Readers of Short Book Reviews will be disappointed at the lack of a fundamental role for stochastic ideas, i.e. uncertainty, error and sampling. Probabilistic models are introduced where unavoidable, in genetics and epidemics, but readers are given the misleading advice that stochastic effects usually do not matter when the population size exceeds thirty.

Contents:
1. Introduction
2. Design and bioavailability studies
3. Statistical inference for effects from a standard2x2 crossover design
4. Statistical methods for average bioavailability
5. Power and sample size determination
6. Tranformation and analysis of individual subject ratios
7. The assessment of inter- and intra-subject variabilities
8. Assumptions and outliers detection
9. Optional crossover designs for two formulations
10. Assessment of bioequivalence for more than two formulations
11. Assessment of bioequivalence for drugs with negligible plasma levels
12. Some related problems in bioavailability studies

Bioavailability and bioequivalence studies play an important role in the drug development process. This text gives a thorough review of the statistical aspects of their design and analysis. It is written for statisticians, primarily those in the pharmaceutical industry who are responsible for these types of studies. The text gives an up-to-date description of the relevant ideas and techniques used for bioavailability and bioequivalence studies, and should be a valuable resource for this audience.

Contents:
1. Introduction
2. Estimating engineering risks
3. Toxicity, toxicology and nutrition
4. Estimation of risk from observation on man
5. Risk perception
6. Risk management
APPENDIX: Costs and benefits of risk reduction

This is in part an update of a 1983 Royal Society report on risk assessment; it also includes new sections on risk perception and risk management. The six chapters are written by different teams of authors and are more or less independent of one another. There is little direct discussion of statistical methods but statistical considerations are prominent in Chapters 3 and 4, both of which had at least one statistician as a co-author. (Sir David Cox in fact co-chaired the group for Chapter 4.) The volume provides an excellent introduction to the main aspects of risk, and a good deal of information on several specific types of risk to humans.

Contents:
1. Basic probability concepts
2. Some initial definitions
3. Some results independent of specific distributions
4. Discrete distributions and combinatorials
5. Specific discrete distributions
6. The normal (or Gaussian) distribution and other continuous distributions
7. Generating functions and characteristic functions
8. The Monte Carlo method: Computer simulation of experiments
9. Two-dimensional and multi-dimensional distributions
10. The central limit theorem
11. Inverse probability: Confidence limits
12. Methods for estimating parameters. Least squares and maximum likelihood
13. Curve fitting
14. Bartlett S function; estimating likelihood ratios needed for an experiment
15. Interpolating functions and unfolding problems
16. Fitting data with correlations and constraints
17. Beyond maximum likelihood and least squares; robust methods

The purpose here is to introduce students working in experimental physics to statistics and probability and to present some more advanced techniques that the author has found useful in his own work. Inevitably this leads to a certain unevenness as the author needs to discuss sophisticated methodology but has not the space necessary to develop the ideas in a logical way. A further difficulty is that explanations are often unclear and poorly motivated; some symbols are not defined and there are occasional loose or vague statements such as "many distributions approach the normal distribution in some limit of large numbers". Any physics student hoping to learn relevant probability and statistics would need to come equipped with a good background in those very areas to be able to fill in the gaps in the text and to get on top of the material.

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

Readership: Health professionals, introductory biostatistics students, lecturers

"The greatest danger is in statistical analysis untouched by the human mind." One has to like a book with this quote on page 3 !
This large volume is designed as an introductory text for individuals with no statistical training nor mathematical background through algebra. It is written by two individuals with a wealth of personal experience in biostatistics and an obvious desire to convey their personal pleasure in the subject to others. The rather unique last chapter is only one example of this. The relative attention given to dif-ferent topics and the implicit foundational approach may be questioned but the problems and examples will, by themselves, make the book of considerable use to many instructors.
The authors indicate that the material presented is often difficult and challenging. For individuals with the presumed background, it will be heavy going at times and, in general, I think many instructors will have to carefully tailor the order and level of presentation. For example, log linear models appear on page 263 as the first statistical method involving more than two variables. Simple linear regression appears on page 352 and logistic regression not until page 631.
Nevertheless, whether used as a primary course text or not, this book should prove a valuable teaching resource.

Contents:
1. Introduction
2. The basic process capability indices: CP, Cpk and their modifications
3. The Cpm index and related indexes
4. Process capability indices under non-normality: Robustness properties
5. Multivariate process capability indices
Notes on computer programs
Postscript

Readership: Statistical researchers curious about what's happening in industry, and practitioners in industry who need theoretical support in dealing with these controversial process capability indices

This monograph is an important and very welcome addition to the literature on process capability indices which have aroused much controversy in manufacturing practice during the past decade. Zealous industrial customers have demanded these statistics of their suppliers to prove the capability of their manufacturing processes lest their business be taken else-where; while statistically sophisticated suppliers have resisted the use of these one-dimensional measures to represent something as complex as "capability". Not the least of the heart-stopping abuses of these indices has been their use as deterministic quantities on which to base business decisions and process control decisions, while the core component of all the indices is a very random and possibly biased sample standard deviation calculated from production data. This book gives a rigorous statistical foundation to the capability in-dices currently in most frequent use, bridging the gap between theoreticians and practitioners. Because of its emphasis on theory, the book will not be accessible to most managers, but it will suitably arm the statisticians who must face these indices and their variability in their work. The book is fairly comprehensive in its overview of the literature in the field through early 1993. Taking a cue from the Foreword by R. Dovich, I would hope that this book will help moderate the debate on process capability measures by enabling discussion to proceed on a more informed basis, in order that industry can get on with real quality improvement rather than focusing on a single, possibly misleading, index.

Contents:
1. Introduction
2. Foundations of statistical quality assurance
3. Acceptance sampling
4. Statistical process control
5. Statistical tables
6. Solutions to exercises
7. Index lists

Readership: Students, scientists and professionals involved in quality assurance

Intensified emphasis on the issue of quality as a prime factor in business competition recently led to a revival in studying statistical quality assurance methods. In contrast to many recent low-level publications, this book provides readers with a complete introduction to acceptance sampling and process control methods, at the same time containing sufficient mathematical motivation and avoiding unnecessary abstraction. Mathematically oriented readers will like the fact that at last a text appears which provides a sound understanding of the basic statistical theory involved. Quality assurance conceptualists will criticize the presence of good old acceptance sampling methods and the absence of experimental design methods as advocated and developed in the so-called Taguchi methods.

Contents:
1. Introduction
2. A simple factorial grazing experiment measured on27 occasions
3. A 3x3 factorial with quantitative levels
4. Definable within-individual comparisons
5. Covariance
6. Pre-treatment observations in the design of experiments
7. Weighted regression, goodness-of-fit and related topics
8. Environmental variables
9. Correlation between series of random variables
10. Response (reaction) times

Readership: Research workers who are interested in the analysis of sequential data

This book introduces research workers to a two-stage method of analyzing data from comparative experiments with sequential observations and demonstrates special features of the design of such experiments. The book contains examples in the area of agri-cultural experiments. It provides step-by-step procedures to guide the reader in the analysis and the construction of appropriate matrices used in regression models at various stages. Since the text is probably written with the intent of being a practical guide for research workers, very little theoretical justification is given. The first few chapters deal with the most common problem of treatment comparisons. The later chapters gradually introduce the idea of controlling for confounding effects by blocking factors and covariates. The last two chapters introduce a simple ex-ample of bivariate sequential data and the analysis of reaction time outcome instead of sequential data. The author mentions very briefly, one paragraph, alter-native approaches in the bibliographical note. It might be more helpful to the reader if an in-depth survey of the general approaches for sequential data analysis and a comparison of different methods had been included in the bibliographical note.

Contents:
Introduction
1. Univariate data
2. Bi- and multivariate data
3. Ancillary techniques
4. Complex examples
5. Appendices

This book is written by two analytical chemists, whose stated aim is to give a cookbook of statistics for personal computer users which addresses examples that a chemist would understand and which includes techniques such as linear regression without what they call the idiosyncratic nomenclatures of introductory courses in statistics. It comes with a computer disc of data files and BASIC programs. There is no non-parametric statistics and no time series analysis. The statistical problems of analytical chemistry deserve a more reasoned treatment than that given here.
The book fails to convey the way a statistician would examine and test evidence; in particular it does not stress the need to consider assumptions before applying tests. For example, to compare means when variances differ we are told that opinions diverge as to the propriety of a t-test, but in practice a t-test will be performed. No alternatives are discussed. The closest I could get to finding out from the book what is understood by correlation was that the correlation coefficient is a statistical measure found in most software packages. Worse, it is misleading. We read that, theoretically, observations must be drawn from a normal population for the standard deviation to be used as a measure of dispersion; that the probability density is the expected frequency of observation; that the expected value is either a theoretical value or an experimental average; that the F-test answers whether the sample standard deviation is unbiased ("Does the found standard deviation correspond to expectations?"). A value predicted from an estimated regression is called the probable value.

Contents:
PART 1: Theoretical Bases of Population Dynamics
1. Basic properties and structure of population processes
2. Structures and patterns of population processes
3. Statistical analysis of population fluctuations
4. Population process models
PART 2: Analysis of Classic Cases
5. Analysis of the lynx 10-year cycle
6. Snowshoe hare demography
7. Density effects on the dynamics of a single-species population: Utida's experiments on the azuki bean weevil
8. Dynamics of a host-parasitoid interaction system: Utida's experimental study
9. Dynamics of the spruce budworm outbreak process

The first half of this book, Part 1 is devoted to the development of simple 1-species and 2-species population models in discrete time, and to techniques for analyzing population time series to discover how they are regulated. The mathematical treatment is simple but is used carefully to lend precision to bio-logical concepts such as density dependence and persistence. The statistical methods are also simple, consisting mainly of the use of regression, autocorrelation and partial autocorrelation to analyze population fluctuations. Simulation is used extensively to demonstrate model behaviour and the ability of the statistical methods to reveal them through the obscuring mists of non-systematic noise.
The applications of these methods in Part 2 are unusual in their thoroughness and fascinating in their sharpness of insight. Royama demonstrates the need to examine all available data on environmental and demographic influences on every aspect of reproductive and survival rate. Yet he also shows that simple models will often suffice where more complex biological pro-cesses were once thought necessary; an example is the Moran effect to explain inter-regional synchrony.

Contents:
1. Introduction
2. Preliminaries
3. Some common probability distributions
4. Distributions related to the Gaussian
5. Sampling
6. Distributions of sample statistics
7. Bayes' theorem and maximum likelihood
8. Confidence intervals
9. Hypothesis testing
10. The theory of maximum likelihood estimators
11. Least squares fitting for linear models
12. Hypothesis testing in the linear model
13. Rank correlation coefficients
14. Tests of fit
15. Robust tests for means
Epilogue
Some numerical exercises

Readership: "Aimed at a diverse scientific audience, including physicists, astronomers, chemists, geologists, and economists"(book cover)

Essentially, this short book provides useful mathematical underpinning to a range of statistical procedures and concepts. Readers need to have encountered many of these already if they are to understand the text. Thus the claim on the cover that 'A reader without previous exposure to statistics will finish the book with a sound working knowledge of statistical methods' is unrealistic. Merits include the reader-friendly style, and the inclusion of such topics as bootstrap and jackknife. Demerits include a somewhat cavalier attitude to various statistical concepts, for example, "[?-1.96ó, ?+1.96ó] is a 95% confidence interval for (a Gaussian) observation"; significance tests are carried out at (say) the "95% confidence level", yet a "test of size á" suddenly appears without explanation. The ninety problems with detailed answers, occupying about a third of the book, are a valuable feature. Lecturers needing supporting material for an already-planned statistics course seem a suitable readership.

Contents:
1. Introduction
2. Cohort design
3. Systematic error
4. Random error
5. Exercises for Chapters 1-4
6. Principles of case-control design
7. Case-control Design A
8. Case-control Design B
9. Case-control Design C
10. Exercises for Chapters 6-9
11. Choice of study design
12. Interpretation of results

The text is lean (88 pages), the exercises/ answers muscular (44 pages) and the references pleasantly plump (40 pages, 700 citations). The stated goal of this volume is to coach the reader in critical assessment of epidemiologic study, design and accuracy. A series of definitions and examples describes the methods of epidemiology. The text is in fact largely an amplification of the six-page glossary. Statistical formulae are completely absent which frustrated this reviewer. However, it is entirely possible that medical scientists will like the omission of statistical techniques.

Contents:
1. Introduction
2. Functions of survival time
3. Examples of survival data analysis
4. Nonparametric methods of estimating survival functions
5. Nonparametric methods for comparing survival distributions
6. Some well-known survival distributions and their applications
7. Graphical methods for survival distribution fitting and goodness of fit tests
8. Analytical estimation procedures for survival distributions
9. Parametric methods for comparing two survival distributions
10. Identification of prognostic factors related to survival time
11. Identification of risk factors related to dichotomous data
12. Planning and design of clinical trials I
13. Planning and design of clinical trials II

This is a revised edition of a book first published in 1980. According to the author's preface "The book has been written for biomedical investigators, statisticians, epidemiologists and researchers in other disciplines who are involved or interested in analyzing survival data". In addition to standard topics in survival analysis such as life-table estimation, proportional hazards models and the fitting of parametric survival curves, the book includes discussion of logistic regression, discrimination, and topics in the planning of clinical trials. The book is intended to be less mathematically demanding than texts such as Kalbfleisch and Prentice (1980) [Short Book Reviews, Vol.0, p.2] or Cox and Oakes (1984) [Short Book Reviews, Vol.4, p.35], and the treatment is more dis-cursive than in those books. However, some degree of mathematical sophistication is required.

Contents:
1. The randomized controlled clinical trial
2. Probability models and clinical medicine
3. Significance tests versus hypothesis tests
4. Neyman's insights
5. Reconciling Fisher and Neyman
6. Continuous measures taken over time
7. Combining data across measures
8. Counts of events
9. Permutation tests and resampling techniques
10. Neyman's restricted chi-square tests
11. Miscellaneous methods for increasing power
12. Bayesian estimation
13. An example of the type of analysis suggested in this book

For those who know or know of David Salsburg the most radical attribute of this book is that it is not radical. Salsburg does an excellent job in his personalized and practical summary of some major statistical controversies of the 20th Century. Though this summary relates only indirectly to his principal thesis of "statistical science, please", my paraphrase, it is entertaining and informative. Salsburg's view from the trenches leads to an approach to hypothesis testing based on targeted, restricted, alternatives instead of omnibus tests. He proposes that these alternatives must be based on scientific understandings. If aimed in the appropriate direction restricted alternatives will increase power. He explains his philosophy through creative approaches to interesting examples and includes extensive data and analysis in Chapter 13.
I am not completely pleased with the book. Salsburg's historical review adds little to recent books. He tantalizes the reader until finally on page 29 he mentions "Bayes". His examples, while realistic, are from applications where sufficient information exists to formulate a restricted alternative. At times his tone is a bit too cute.
Though experienced statisticians may react to the book by thinking that they already know and practice a similar philosophy, all readers will benefit from the reminder that most hypothesis tests target a specific alternative and that, as Salsburg writes on page 160: "...statistics should be the handmaiden of medical research ... and not its jailer."

Contents:
1. Introduction
2. General concepts of probability
3. Random variables, probability distributions, and characteristic functions
4. Stochastic convergence and limit theorems
5. Concepts of statistics
6. Univariate distributions
7. Multivariate distributions
8. Order statistics and related distributions
9. Statistical inference: Parametric point estimation
10. Testing of statistical hypotheses
11. Large sample methods
12. Statistical decision theory
13. Sequential analysis
14. Nonparametric methods
15. General linear hypothesis and analysis of variance
16. Some applications of analysis of variance

Readership: Undergraduate and conversion postgraduate students in probability and statistics

This book, an up-date of the first edition of approximately eighteen years ago, provides a clear and solid foundation for the current teaching of probability and theoretical statistics within the classical framework. Contributions of alternative approaches are not considered. In particular, subjective probability and Bayesian inference do not appear except, and with-out discussion, in a very classical approach to statistical decision theory. A thorough knowledge of calculus is assumed and, in view of current student inabilities to manipulate matrices, there is a welcome appendix on vectors and matrices.

Contents:
1. Scope and method of the study of general social attitudes
2. Methods in comparison
3. On the content of questions
4. On the nature of the samples
5. The position of groups
6. Comparison by means of the effect of attributes
7. Macro-analysis by cohort analysis
8. Comparisons by means of systems of thought
9. Design and comparison of questions
10. Inquiries through question design changes
11. Expansion of the cultural link analysis of comparative research
12. Total grasp based on dissimilarity of subjects
13. Methods for intensifying information
14. Cohort analysis methods
15. A method for seeking out the system of thought

This monograph deals with the problems connected with the comparison of results of attitude surveys conducted in different countries. The authors point out the difficulties they encountered in conducting and analyzing results from surveys in Japan, Hawaii, Southeast Asia, Philippines and France. The authors note the important role of wording the quest-ions and the way they affect the response. The main reason for the difficulty is the differing cultural backgrounds. The monograph describes various ways in which the effect can be minimized. This volume is a valuable contribution in an increasingly important area.

Contents:
1. Introduction
2. Differential equations models
3. Difference equations models
4. Delay differential equations models
5. Age and internal structure

Readership: Modelers, epidemiologists, population geneticists, ecologists, applied mathematicians

This book provides a mathematical study of the transmission of infectious diseases in which one mode of transmission is from parent to offspring. A deterministic approach is taken and models are formed in terms of differential or difference equations, some-times including delay terms. The final chapter considers age structure and involves partial differential equations. No attempt is made to consider stochastic variations. Modeling revolves around equations for the number of susceptible, exposed, infected, and removed individuals and does not assume a fixed population size. Models related to specific diseases are discussed briefly.

Contents:
1. Introduction
2. Basic mathematical and statistical concepts
3. Aims, ideas and models of factor analysis
4. R-mode methods
5. Q-mode methods
6. Q-R-mode methods
7. Steps in the analysis
8. Examples and case histories

The unwary may easily be misled by the title; "factor analysis" is used here informally, to signify the statistical application of the algebra of eigen-values and eigenvectors to the analysis of single multivariate samples. Thus a wide range of techniques, for example principal components, principal coordinates, correspondence analysis, are described in addition to factor analysis in its more usual sense, distinguished here as "true" factor analysis. The book is actually a second edition of the same authors' Geo-logical Factor Analysis of 1976; it has been updated by the inclusion of recent methodology, for example analysis of compositional data, assessment of influence and stability of factors, analysis of asymmetry and its scope has been widened to all the natural sciences with a long chapter describing many varied case studies, and there is an extensive appendix by L. Marcus of MATLAB programs implementing most of the discussed techniques. Statisticians will find some unfamiliar slants on fami-liar methods, and all applied readers should find plenty to interest them.

Contents:
1. Introduction
2. The fundamental problem of pattern recognition
3. Images, filter images, and shapes: The topological framework
4. Objects or images, structures or patterns, classification and recognition: The structural or syntactic framework
5. General formalization of the pattern recognition problem: The categorical framework
6. Conclusion

The flavour of this book is well summarized by the cover blurb: '[This book] addresses the field of pattern recognition from a basic mathematical view-point - treating algebraic, topological, and categori-cal approaches of pattern recognition using concepts from homotopy, shape, and fiber space theories as well as computational topology.' It is thus essentially a mathematics text and is abstract and theoretical. There are no sets of data in it. Indeed, the author does suggest the alternative title of 'The Mathematical Challenge in Pattern Recognition.' The first edition (1989) [Short Book Reviews, Vol. 9, p.26] has been extended by the inclusion of additional material, notably five appendices showing relationships between some of the mathematical areas involved and a new selection of fiberwise topology.

Contents:
Introduction
1. Introduction to multivariate space (P.J. Lewi)
2. Multivariate data display (P.J. Lewi)
3. Vectors and matrices: Basic matrix algebra(N. Bratchell)
4. The mathematics of pattern recognition (N. Bratchell)
5. Data reduction using principal components analysis(J.M. Deane)
6. Cluster analysis (N. Bratchell)
7. SIMCA - classification by means of disjoint cross-validated principal components models (O.M. Kvalheim and T.V. Karstang)
8. Hard modelling in supervised pattern recognition(D. Coomans and D.L. Massart)
Software Appendices
SPECTRAMAP (P.J. Lewi)
SIRIUS (O.M. Kvalheim and T.V. Karstang)

Chapter 1 begins with a conventional objects-by-variables data matrix and describes geometrical ways of viewing operations on this matrix, including multiple regression, projection, rotation, and principal components analysis. Chapter 2 describes data displays obtained via some form of factor analysis, namely correspondence analysis, principal components analysis, and spectral map analysis. Associated software, SPECTRAMAP, described in the appendix, can be obtained to produce these displays. Chapter 3 covers topics such as trace, rank generalized inverse, and so on, and this work is extended in Chapter 4, relating the algebra and geometry to operations used for the statistical techniques. Chapters 5 and 6 deal with the respective topics of principal components analysis and cluster analysis. The former includes an extensive discussion of 'determining the true dimensionality of the data'. Chapter 7 describes an approach to clustering the points according to their distance from the best fit-ting lines to the clusters. Chapter 8 describes supervised pattern recognition, formulating a classification rule using a sample of classified cases.
Different authors in this book use different notation, and some topics, such as singular value decomposition and data pre-treatment for example, are introduced several times in different ways. While appreciating that readers may, at first, find this confusing, the editor suggests that this can be useful in that people find different approaches to the same topic complementary: if one treatment seems inaccessible perhaps another will be easier. Certainly, there is much cross-referencing between different sections.
Each chapter has a number of questions scattered throughout the text, with answers given at the end and, as befitting a book on an applications area, there are many examples.
The book does not go very deeply into the subject matter. Nevertheless, it is easy to read and would, I think, be useful as an introduction to certain pattern recognition topics for a wider cross-section of readers than merely chemometricians.

Contents:
1. Introduction
2. Development of Pitman's measure of closeness
3. Anomalies with PMC
4. Pairwise comparisons
5. Pitman-closest estimators
6. Asymptotics and PMC

This research monograph assembles the wide-spread material concerning Pitman's measure of closeness (PMC) that is available in the literature and much of which is not widely known. The PMC, introduced by Pitman in 1937, compares two estimators 01 and 02 by the probability that |01 - 0| is less than |02 - 0|.
As indicated by the subtitle, "A comparison of statistical estimators", the authors recommend Pitman closeness as an interesting alternative criterion for comparing estimators. They investigate the usefulness of the PMC for this purpose, and discuss the properties of "Pitman-closest" estimators. This is done both from a frequentist and Bayesian point of view, in both small-sample and large-sample settings.
The book contains many fascinating examples and results. The reader's reaction to this material is likely to be strongly influenced by his feeling about the property that has made the PMC controversial. Its intransitivity, which in the comparison of three estimators, allows it to prefer estimators A to B, B to C and C to A. The authors face this issue in Chapter 3 in which they try to deflect the objection by pointing to other situations that involve intransitivity, including a brief discussion of Arrow's Impossibility Theorem. While it is good to have such an alternative measure available, this reviewer remains unconvinced and prefers criteria not suffering from this defect.

Contents:
1. Introducing nonparametric methods
2. Location tests for single samples
3. Rank transformations and other tests for single samples
4. Methods for paired samples
5. Methods for two independent samples
6. Three or more samples
7. Correlation and concordance
8. Regression
9. Categorical data
10. Association in categorical data
11. Robustness
12. Other developments

This text retains from the first edition its low technical level and its emphasis on methods and applications [Short Book Reviews, Vol. 9, p.47]. The major changes are an added chapter on each of regression and categorical data, plus added emphasis on the use of software for exact permutational computations. For instance, the author shows how to use StatXact to unify nonparametric rank procedures and certain analyses for contingency tables. Other positive features include strong emphasis on estimation as well as testing, introductions to recent developments, and many "real-data" examples and exercises.

Contents:
1. The geometry of exponential families
2. Calculus on manifolds
3. Statistical manifolds
4. Connections
5. Curvature
6. Information metrics and statistical divergences
7. Asymptotics
8. Bundles and tensors
9. Higher order geometry

Statistical asymptotics is notorious for leading to complicated calculations and formulae. One can make some sense of the results by using ideas from differential geometry. However, the basic geometry of Riemannian metrics is not adequate; one has to come to terms with the theory of connections and their curvature, and even more specialized ideas relating to geometrization of Taylor Series. This book does an excel-lent job of explaining these geometrical ideas from a statistical point of view, using statistical examples as opposed to, for example, dynamical examples to illustrate the geometric concepts. It should prove most helpful in enlarging the typical statistician's working geometric vocabulary.

Contents:
PART I : An Introduction to Point Processes
1. Strong approximations
2. Poisson and Cox processes
3. Densities and distances
PART II : Point Processes in Action
4. Nonparametric curve estimation
5. Sampling from finite populations
6. Extreme value models
7. Image restoration, spatial statistics
PART III: An Outlook on Further Important Approaches
8. Weak approximation
9. Counting processes and martingales

The organization of the book in three main blocks reflects the author's main intention of providing a course on point processes where theory as well as applications are both fairly treated. Though using the more measure-theoretic language, now well established and used by specialists, the main thread of the book never is far away from useful applications of the theory. The mathematically more sophisticated techniques, like Palm and Campbell measures, topological considerations, are only introduced when enough motivation has been given to justify their discussion. Numerous exercises are provided. I am sure that, as a graduate textbook, this course will provide a useful alternative to the growing list of textbooks on the subject.

Contents:
1. Preliminaries
2. Sensitivity analysis and stochastic optimization of discrete event static systems
3. Sensitivity analysis and stochastic optimization of discrete event dynamic systems
4. What is a "good" reference system to simulate
5. Extensions of the SF method
6. Statistical inference of stochastic optimization programs

Readership: Students and researchers in operational research, management and industrial engineering, statistics and computer science

The behaviour of complex stochastic systems like queues or networks may be studied through simulation. Of interest are the sensitivity of the system to changes in parameters, and optimization of measures of performance. Statisticians might approach these is-sues by running several simulations, for different parameter combinations. This book advocates a different procedure, using just one simulation. In outline, differentiating expected measures of performance provides measures of sensitivity. If the order, of integration and differentiation, can be interchanged, then score functions result, which can be estimated from a single simulation. Approximate solutions to complex optimization problems build on this work, providing stochastic analogues to deterministic problems, and then solving the former by mathematical programming.
This book is based on a graduate course. The authors have tried to emphasize concepts, rather than detail, but the reader requires far more than the suggested "basic knowledge of probability, statistics and optimization" in order to follow all of the development.
This is an important, fascinating book. It should be studied carefully by anyone contemplating a large-scale computer simulation. Sadly the index barely stretches to two pages.

Contents:
1. Philosophical foundations
2. Maximum entropy estimates of the density of the earth
3. Maximum entropy geotomography (Boltzmann method)
4. Maximum entropy tomography (Boltzmann method)
5. Maximum entropy spectral analysis (Gibb's method)
6. Maximum entropy coherent spectral analysis(Boltzmann method)
7. Maximum entropy photon image restoration
8. Reliability-risk maximum entropy analysis
APPENDIX A: Derivation of the Gibbs ME distribution for a spherical earth density
APPENDIX B: Derivation of the Gibbs ME distribution for laterally varying earth density
APPENDIX C: Statistical distribution of a real-valued parameter by the Darwin-Fowler method
APPENDIX D: Derivation of the maximum entropy complex multivariate Gaussian probability density
APPENDIX E: The entropy of failure probabilities of a number of independent system components

Readership: Engineers, applied statisticians especially in engineering, image analysis and the earth and space sciences

The maximum entropy method has its origin in statistical mechanics where it was developed by two, slightly different, approaches due to Gibbs and Boltzmann, respectively. In the sixties and seventies it gained wider prominence as a procedure for the handling of 'inverse data problems' in physics and geophysics, through the works of Jaynes and Burg. The book under review briefly sets out the main points of the method in the introductory chapter and then presents rather detailed discussions of a variety of interesting and successful applications of the method, one chapter being devoted to each of the applications. The subject problems concerned are from geophysics, engineering and image analysis. The problems are mostly quite complex and it is a strength of the book that they are discussed fairly thoroughly. Even so, however, the background technicalities are such that most statisticians will have difficulty in following the developments closely.
The author is a self-confessed devotee of the maximum entropy method, especially the Boltzmann version. As part of his effort to convince others he presents in the introductory chapter a 'dialogue' between a maximum entropy analyst and what is purported to be a 'classical statistician'. Unfortunately, though not surprisingly, 'Mr. Classi' is but a caricature statistician, and no real contest emerges.

Contents:
1. Introduction
2. Stationary stochastic processes
3. Deterministic spectral analysis
4. Foundations for stochastic spectral analysis
5. Linear time-invariant filters
6. Nonparametric spectral estimation
7. Multitaper spectral estimation
8. Calculation of discrete prolate spheroidal sequences
9. Parametric spectral estimation
10. Harmonic analysis

In this book the authors consider various parametric and non-parametric methods of estimation of the spectral density function where a sample from a stationary time series is available. The emphasis is on the "multitaper" method of estimation, where one uses several orthogonal tapers for estimation of the spectral density function from the same data, and the final estimate is obtained by an average of these estimates. The treatment is thorough and exhaustive. The attractive feature of the book is it contains several illustrations taken from meteorology, oceanography and many other disciplines. It is an extremely
useful book, and I strongly recommend it to all who are interested in spectral analysis.

Contents
1. General concepts of stochastic processes
2. Analytic theory of Markov chains
3. Properties of sample functions
4. Some topics in Markov chains
5. Basic theory of birth and death processes
6. Construction theory of birth and death processes
7. Analytic construction of birth and death processes
8. Bilateral birth and death processes
APPENDIX 1: Excessive functions of Markov chains with discrete time
APPENDIX 2: -systems and the -system method

Large portions of this highly-technical and scholarly-written monograph have appeared in 1980 in a Chinese edition. The first half of the book deals with general Markov chains and brings an update of Chung's classic textbook Markov Chains with Stationary Transition Probabilities. The second part gives a very detailed and neat analytical development of birth and death processes; a major part of this material appears for the first time in a translated form. The general tone of the book is a bit dry. Nevertheless, the wealth of included material as well as the streamlined approach deserves our greatest admiration.

Contents:
1. A preview
2. Geometric moment theory
3. Moment problems of Kantorovich type and Kantorovich radius
4. Moment problems related to c-rounding proportions
5. The Lévy radius
6. The Prokhorov radius
7. Probability measures, positive linear operators and Korovkin type inequalities
8. Optimal Korovkin type inequalities under convexity
9. Optimal Korovkin type inequalities for convolution type operators
10. Optimal Korovkin type inequalities for positive linear stochastic operators
11. Optimal Korovkin type inequalities for positive linear operators using an extended complete Tchebycheff system
12. A general Ê-attained inequality related to the weak convergence of probability measures to the unit measure
13. A general stochastic inequality involving basic moments
14. Miscellaneous sharp inequalities and Korovkin type convergence theorems involving sequences of basic moments
15. A discrete stochastic Korovkin type convergence theorem

Readership: Researchers in probability, approximation theory, numerical analysis, statistics, functional analysis

"The book is a research monograph giving an account of the author's work of the last eleven years in the field of applications of geometric moment theory to probability (Moment) Theory and especially Approximation Theory." Moments and their estimates are daily bread and butter in probability theory and statistics. The author gives a comprehensive and self-contained survey of methods and results in geometric moment theory. He poses and solves several basic moment problems, for example those of Kantorovich type. He also deals with the problems of weak convergence of finite measures to the unit measure and the pointwise convergence of positive linear operators to the unit operator.
This book will be of great use for all re-searchers who are basically interested in the inter-relationships between analysis and probability theory. Although being addressed to the specialist the mono-graph is of interest to all those who wish to obtain an overview of the results and techniques in this field.

Contents:
1. Deterministic optimal control
2. Viscosity solutions
3. Optimal control of Markov processes: Classical solutions
4. Controlled Markov diffusions in Rn
5. Viscosity solutions: Second-order case
6. Logarithmic transformations
7. Singular perturbations
8. Singular stochastic control
9. Finite-difference numerical approximations

Applications of controlled Markov processes can be found in many fields including manufacturing, economics, communication systems and management. This book gives a comprehensive treatment of continuous time Markov processes and the associated optimal stochastic control problems. The key tool used is that of dynamic programming. This leads to the Hamilton-Jacobi-Bellman partial differential equation. However, in general, the value function for stochastic optimal control problems is not smooth enough to satisfy this equation in the usual way. Thus a weak form of solution is needed. This is provided by a relaxation method commonly known as 'viscosity' solutions. The book gives a complete treatment of this theory including recent extensions and embellishments. The book is recommended to anyone who has an interest in this important class of control problems.

Contents:
Introduction
PART I : Essentials of Nonlinear Dynamics
PART II : Essentials of Information Theory and Thermodynamics
PART III: Thermostatics of Multifractals
PART IV : Dynamical Analysis of Chaotic Systems
PART V : Advanced Thermodynamics

Even very low-dimensional, simple deterministic systems can exhibit an unpredictable, quasistochastic long-time behaviour ['chaos'] characterized by sensitive dependence on initial conditions. In order to describe the important features of such systems, many concepts have been borrowed from thermodynamics and statistical mechanics - with great effect. This book explains the importance and usefulness of these concepts; this is done in the language of physics, so that the mathematical knowledge required of the reader is by no means excessive.

Contents:
1. Introduction
2. Asymptotic properties of weighted empiricals
3. Linear rank and signed rank statistics
4. M, R and some scale estimators
5. Minimum distance estimators
6. Goodness-of-fit tests for the errors
7. Autoregression

For a random sample from a distribution know-ledge of the empirical distribution function enables one to recover the observations up to a permutation. For this and other reasons the empirical distribution function has been the subject of much study, and has been used as the direct basis of a number of tests of hypotheses. Suppose instead one has n pairs (xi, yi), where xi is a vector, in a regression situation: what then? There is a strong argument for saying that the natural object is a weighted empirical distribution function, of the residuals, where the weights are the xi; if one considers autoregression one also finds a need to deal with random weights. The present monograph discusses the properties of weighted empirical distribution functions and then their applications to various problems of estimation or testing regression parameters, linear rank statistics, M-estimators and R-estimators, minimum distance estimators, and of goodness-of-fit of the residual distribution.
It is very comprehensive and, once past the introductory chapter sections, very technical.

Contents:
Introduction
1. Fourier series and integrals, filtering and sampling
2. Multiresolution approximations of L2(Rn)
3. Orthonormal wavelet bases
4. Non-orthogonal wavelets
5. Wavelets, the Hardy space H1 and its dual BMO
6. Wavelets and spaces of functions and distributions
New references on wavelets and their applications

Readership: Mathematicians, statisticians and engineers working on signal-processing and image- processing

This book represents the English language translation of an excellent introduction to wavelets by a leading researcher in this field.
Wavelets extend and simplify Fourier methods in analysis. Unlike Fourier series, the wavelet series translates many properties of the function or distribution simply and precisely. In particular, wavelets are well-suited for analyzing transient behaviour in a function or distribution.
This book constitutes volume one of the English language translation of Yves Meyer's classic, Les Ondelletes. This lucid account of the theory behind wavelets is aimed at the postgraduate level. The author takes pains to motivate a rigorous presentation. There is an extensive bibliography; statisticians and engine-ers will appreciate the short, but useful, section containing new references on wavelet applications. Volumes 2 and 3 are intended primarily for mathematicians and describe the use of wavelets in Calderon's programme in operator theory.

Contents:
1. Probabilistic preliminaries
2. Hp(p?1) martingales
3. Ö-inequalities on martingales
4. BMO martingales
5. Martingale transforms
6. Weight theory and weighted Ö-inequalities
7. Regular martingales
8. Some applications of martingale techniques in harmonic analysis

This book gives the mathematical theory of discrete time martingales and various related spaces and inequalities; it also discusses some applications to areas such as harmonic analysis. Unfortunately the English language used is very strange and hard to understand. From the first sentence of the preface to the last of Chapter 8 the constructions are unnatural. It is incredible the book was not read before publication by someone fluent in English. This, together with the technical concentrated subject matter, will make the book of limited appeal and usefulness. In addition there are few references later than the early 1980's; therefore, the book does not include recent contributions of the French school.

Contents:
1. Entropy
2. Noiseless coding
3. Noisy coding
4. General remarks on codes
5. Linear codes
6. Some linear codes
7. Finite fields and cyclic codes
8. Some cyclic codes

The first quarter of the book is devoted to information theory in the Shannon sense and the remainder of the book is devoted to algebraic coding theory. The book is written in order that the section on information theory can be omitted without loss of understanding of the coding theory. The treatment of information theory is at a fairly elementary level. The coding theorem for noisy channels is described for the discrete memoryless channel, but is proved only for the binary symmetric channel. The part on coding theory is more complete and covers a good range of topics, including the usual Hamming, Golay, Reed-Muller, Reed-Solomon, and BCH codes, as well as Justeson codes and Goppa codes. There are performance bounds and some discussion of asymptotic performance. There is no discussion of the circuits which might implement the coding and decoding algorithms. There is a section which describes some of the connections between coding and combinatorial designs.

Contents:
1. Basic ideas in probability
2. The role of entropy: A quantitative measure of information, uncertainty and complexity
3. The principle of maximum entropy
4. Coding theory and codes with a central dogma
5. The source, transmission and reception of information
6. The information content or complexity of protein families
7. Evolution of the genetic code and its modern characteristics
8. The early earth and the primeval soup
9. Did life emerge by chance from a primeval soup?
10. Self-organization of life scenarios
11. Error theories of ageing
12. Information theory and molecular evolution
Epilogue

The author argues quite strongly that Shannon's information theory should be useful in understanding molecular biology. The noiseless coding theorem of information theory is connected with the genetic code. The coding theorem for a noisy telecommunication channel is interpreted for the transmission of genetic messages, with mutations taking the place of channel noise. The first third of the book is devoted to an introduction to probability and information theory, and the last two thirds is devoted to applications to problems in molecular biology.

Contents:
1. Data and models
2. Maximum-likelihood fitting of simple models
3. Extensions and alternatives
4. Extended models for quantal assay data
5. Describing time to response
6. Over-dispersion
7. Non-parametric and robust methods
8. Design and sequential methods

Readership: Senior undergraduate and graduate students in biostatistics, toxicology and pharmacology

This book is a welcome addition to the analysis of dose-response quantal bioassays which are so important in toxicology and pharmacology. It contains eight chapters with over fifty real sets of data as well as two hundred and sixty-seven exercises with solutions to many. Most major computing software, including SAS, GLIM and MINITAB, are described and illustrated. The proofs and derivations are minimal. The chapters on over-dispersion and time-dependent experiments are extra-ordinarily comprehensive. There are over seven hundred recent references. Although this book does not contain analysis of quantitative assays, it contains the best collection and description of all known methods in quantal bioassays.

Contents:
1. Events and their probabilities
2. Random variables and their distributions
3. Discrete random variables
4. Continuous random variables
5. Generating functions and their applications
6. Markov chains
7. Convergence of random variables
8. Random processes
9. Stationary processes
10. Renewals
11. Queues
12. Martingales
13. Diffusion processes

Already in my 1983 review of the first edition of the book [Short Book Reviews, Vol.3, p.5.], I praised this text as an interesting addition to the long list of textbooks on probability. The present text has an extra two hundred pages which is reflected in changes and updates in all chapters, together with the addition of a new chapter on martingales and the split-ting of a previous chapter on stationarity and diffusion into two separate chapters. Many new examples and exercises have been added. As it stands, this is de-finitely one of my favourites as a textbook. In the self-contained companion volume, all problems and exercises, six hundred and seventy-five, of the main text are listed and solved. This results in a wealth of interesting teaching material at all levels. My main concern remains to what extent students will have the willpower to follow the authors' advice in the preface: 'Make two serious attempts at each question before reading its solution.' It will definitely force examiners to be even more inventive in setting future examination papers.

Contents:
PART I
1. Preliminaries of probability
2. Finite sample spaces
3. Two or more events
4. Random variable, distribution function, and expected value
5. Functions of a random variable
6. Two or more random variables
PART II
7. Statistics
8. Quality control: Control charts
9. Tolerancing, error analysis and parameter uncertainty
10. Reliability engineering, by K.C. Kapur
11. Random processes and congestion
12. Decision trees

Readership: Advanced undergraduate, beginning graduate engineering students; practising engineers

Part I of this book may be thought of as a standard introduction to probability theory for engineers. Most theorems in this part are proved, with the usual exceptions and limitations. Some nice examples, for example in quality control, telephone networks, show the student how he can already use his knowledge at this early stage for solving practical problems. The presentation of the first part is such that it can be used for the practicising engineer as a collection of useful distributions and standard formulas.
Part II of this book is based on the first part but every chapter is self-contained and gives an introduction to various fields of interest of the engineer. The author aims to introduce basic concepts without proofs in applied fields like statistics, quality control and reliability. The interested reader can also find some references to more advanced and specialized topics in these areas thus providing "guidance
needed to get help". These and numerous interesting exercises form the core of this well-written book. I am convinced that many engineering students will find this a very readable introduction to probability theory and its applications.

Contents:
1. Preliminaries: Discrete index sets and/or discrete state spaces
2. Markov chains
3. Renewal theory
4. Point processes
5. Continuous time Markov chains
6. Brownian motion
7. The general random walk

This text is intended to be accessible to beginning students of stochastic processes as well as those with more background; the author indicates some variation in routes through topics according to students' expertise. Though some basic probability theory is required, no previous exposure to measure theory is assumed and, where certain measure-theoretic ideas are introduced, such as in a discussion of stopping times in Chapter 1, over-dependence on them may be avoided. As the contents list indicates, a good variety of topics is treated in some depth, though martingale theory is not included. Chapter 7 deals with Wiener-Hopf factorization, with some applications here, as often elsewhere in the book, to queueing processes. There is a large number of exercises and, frequently characterful, examples. It appears that a companion disc is in preparation for solution of numerical examples.

Contents:
6. Contingency tables
7. Multivariate distributions, inference, regression and canonical correlation
8. MANOVA, discriminant analysis and qualitative response models
9. Principal components, factors and correspondence analysis
10. Cluster analysis and multidimensional scaling

This is the second of a two-volume survey totalling 1352 pages, under the general heading 'Applied Multivariate Data Analysis'. For a review of the first volume, see Short Book Reviews, Vol. 12, p.8. This volume constitutes a thorough treatment of multivariate topics, with a heavy emphasis on working through all the practical details associated with each of the sociological or business types of example chosen to demonstrate the topics. The notation of these examples assumes the use of a computer package, with MINITAB, SPSSX and SAS having been used by the author,
although no specific programming assistance is given. The exercises are follow-ups to these examples, and the twenty-two sets of data are supplied on a disk sup-plied with the book. The questions are designed to extend the theory but are laid out in a helpful manner for the student reader. As with the first volume, the graphics and presentation are good.

Contents:
1. Introduction and preview
2. Basic distribution theory
3. Discrete order statistics
4. Order statistics from some specific distributions
5. Moment relations, bounds, and approximations
6. Characterizations using order statistics
7. Order statistics in statistical inference
8. Asymptotic theory
9. Record values

Readership: Students in mathematical sciences at the first-year graduate and advanced undergraduate levels

Order statistics is perhaps no longer such an important field, at any rate in statistical inference, as was once thought it would be. Only extreme values flourish. Still, any serious probabilist and statistician should know something about the underlying theory, and for such a purpose this book, or part of it, will serve excellently. It is a matter of taste whether the material on characterizations should have been included. 'Record values' is perhaps not always considered as a part of order statistics, but the inclusion herein is logical and welcome.

Contents:
1. Basic ideas and principles
2. Optimal predictors of population quantities
3. Bayes and minimax predictors
4. Maximum likelihood predictors
5. Classical and Bayesian prediction intervals
6. The effects of model misspecification, conditions for robustness, and Bayesian modelling
7. Models with measurement errors
8. Asymptotic properties in finite populations
9. Design characteristics of predictors

Statistical inference for finite populations and survey sampling is introduced from the perspective of superpopulation models. The general approach is to bring to bear classical results in statistical optimality and decision theory in a theory based on predictive inference. The approach adopted emphasizes the optimal-ity properties of particular classes of estimators, for given data, with little discussion of sampling design. Substantial references are provided to the considerable literature in this area, and exercises are provided in
each chapter, making the text suitable for an advanced course in superpopulation models.

Contents:
1. Basic concepts and definitions
2. The Poisson process and its ramifications
3. Elements of Brownian motion
4. Gaussian processes
5. L2 space
6. Second-order processes
7. Spectral analysis of stationary processes
8. Markov processes I
9. Markov processes II: Application of semigroup theory
10. Discrete parameter martingales

Here we have a modern, reasonably rigorous theoretical treatment of some of the major features of the stochastic process landscape. Ideas such as coupling and stochastic integration, familiar in the advanced literature are now making an appearance at a popular graduate level. In particular the differential equation approach to the Poisson process is replaced by methods based around a coupling argument. A lot of effort is devoted to stationary and Markov processes; renewal, queueing and branching processes are discussed only briefly, if at all. The proofs are clear with sufficient detail although an appendix with important results from measure theory would have been useful. Plenty of exercises are given at the end of each chapter. There was very little evidence of the applications referred to in the title.

Contents:
1. Representation and geometry of multivariate data
2. Nonparametric estimation criteria
3. Histograms: Theory and practice
4. Frequency polygons
5. Average shifted histograms
6. Kernel density estimators
7. The curse of dimensionality and dimensionality reduction
8. Nonparametric regression and additive models
9. Other applications

Readership: Graduate students of statistics, statisticians, biostatisticians, electrical engineers, econometricians, and other scientists involved in data analysis

The author's aim is to demonstrate that density estimation can be a powerful tool in spaces with more than two dimensions, with special emphasis on the trivariate and quadrivariate. He attempts to cover both theoretical and practical aspects and hopes the book will serve both as an introductory text and also as a general reference. I think he succeeds.
The development is from the familiar back-ground of the histogram and the relationships between estimators is emphasized. The key role of graphical displays is noted. There is a nice introductory discussion about the merits of different types of method, and how a technique which is theoretically best in some circumstances may in fact not be the one that is practically best. Overall the discussion is well set in historical context and the development from one topic to another is natural. The book is extremely well written. Overall I strongly recommend it to anyone who wants an overview of an area which is of growing importance.

Contents:
PART I : Foundational Issues in Statistical Inference
1. Conditional inference from confidence sets, G. Casella
2. Intervention experiments, randomization and inference, O. Kempthorne
3. Ancillarity, E.L. Lehmann and F.-W. Scholz
4. The Pitman closeness of statistical estimators: Latent years and the renaissance, P.K. Sen
5. Unbiased sequential binomial estimation, Bimal K. Sinha and Bikas K. Sinha
6. Sufficiency, S. Yamada and H. Morimoto
PART II: Bayesian Methods
7. Foundations of statistical quality control, R.E. Barlow and T.Z. Irony
8. Prequential data analysis, A.P. Dawid
9. Bayesian nonparametric inference, T.S. Ferguson, E.G. Phadia and R.C. Tiwari
10. Hierarchical and empirical Bayes multivariate estimation, M. Ghosh
11. Basu's contributions to the foundations of sample survey, G. Meeden
PART III: Sampling, Designs and Majorization
12. Survey sampling - As I understand it, V.P. Godambe
13. Two basic partial orderings for distributions derived from Schur functions and majorization, K. Joag-Dev and J. Sethuraman
14. Optimal integration of surveys, P.K. Pathak and M. Fahimi
15. The model based (prediction) approach to finite population sampling theory, R.M. Royall
16. Sampling theory using experimental design concepts, J. Srivastava and Z. Ouyang

Readership: Statisticians interested in statistical inference, especially its foundations

This volume is a Festschrift to Professor D. Basu on the occasion of this 65th birthday. The six-teen articles that constitute the volume are written by friends and colleagues of Professor Basu. The papers are mostly of the review type, and the majority of them relate closely to Dr. Basu's many, varied and import-ant, contributions to statistical inference.
Overall the volume contains little original material, and some of the papers are rather specialized [4,5,14]. Others [1,3,6] address subject areas of wide interest and basic importance and provide useful over-views of some parts of those areas. However, further parts, of much current interest, are not mentioned or only slightly indicated.
Of very considerable value are the contributions [9] and [16]. The first provides an overview of the extensive amount of material on random probability distributions and their relation to nonparametric Bayesian inference that has appeared within the fifteen years following the 1974 review article by T.S. Ferguson and which is scattered over the statistical and probabilistic literature. Much of the theory revolves around the class of Dirichlet processes. In [16] a highly illuminating and readable discussion is given of the progress made in the model-based approach to finite population sampling and of the light this throws on the randomization-based methodology. It is gratifying to see how basic ideas of main stream statistical inference are coming to play the same kind of role in finite population sampling that they have in other parts of statistics.

Contents:
1. The item characteristic curve: Dichotomous response
2. Estimating the parameters of an item characteristic curve
3. Maximum likelihood estimation of examinee ability
4. Maximum likelihood procedures for estimating both ability and item parameters
5. The Rasch model
6. Parameter estimation via marginal maximum likelihood estimation and an EM algorithm
7. Bayesian parameter estimation procedures
8. The graded item response
9. Nominally scored items
10. Implementation of maximum likelihood estimation of item parameters
11. Implementation of maximum likelihood estimation of examinee's ability
12. Implementation of item parameter estimation via MMLE/EM
13. Implementing the Bayesian approach
14. Implementation of item and ability parameter estimation under a graded response model
15. Implementatin of maximum likelihood estimation of item and ability parameters under normal response scoring

The item response models dealt with in this book belong to the class of factor analysis models with categorical response variables. The author considers only one-dimensional models, and for the most part the special cases where the response is binary. The latter models are widely used in psychometric test construction and analysis, where a set of test item responses is viewed as a two-way table of subjects by items with a response in each cell of the table. This book gives details of the estimation procedures commonly applied to these models with appendices setting out source code programs written in BASIC. There are no data analyses presented, and there is no attempt to provide any substantive justification for the use of such models or the problems associated with the restriction to one-dimension. For readers who want a broader statistical perspective on these models the book by Bartholomew Latent Variables and Factor Analysis can be recommended.

Contents:
PART I : Preliminaries
1. Probability and statistics
2. Probability and stochastic processes
PART II : Stochastic Differential Equations
3. Itô stochastic calculus
4. Stochastic differential equations
5. Stochastic Taylor expansions
PART III: Applications of Stochastic Differential Equations
6. Modelling with stochastic differential equations
7. Applications of stochastic differential equations
PART IV : Time Discrete Approximations
8. Time discrete approximation of deterministic differential equations
9. Introduction to stochastic time discrete approximation
PART V : Strong Approximations
10. Strong Taylor approximations
11. Explicit strong approximations
12. Implicit strong approximations
13. Selected applications of strong approximations
PART VI : Weak Approximations
14. Weak Taylor approximations
15. Explicit and implicit weak approximations
16. Variance reduction methods
17. Selected applications of weak approximations

Readership: Everyone interested in the theory and applications of stochastic differential equations

By now, the theory of stochastic differential equations is well understood. Over the last years, we have seen many important applications in such diverse fields as for instance biology and finance. In this book, the authors give a comprehensive and self-contained overview of the basic stochastic differential equation theory, its many applications and above all, a detailed discussion of the numerical aspects of solving stochastic differential equations. The reader is amply warned that for the latter, one cannot 'just use a stochastic version of the deterministic analogue'. Many examples, exercises and so-called PC-exercises make this text an impressive and scholarly piece of work. The authors do not content themselves with only explaining the mathematical results, but they spend a lot of effort in convincing the reader to actually sit down and work out some examples. More of this hands-on approach is to be expected from the companion volume The Numerical Solution of Stochastic Differential Equations through Computer Experiments. It is really remarkable that the authors have succeeded in writing a text for a very wide audience including those who actually want to solve a given stochastic differential equation and those doing research in the field. It is clear that a short review cannot fully do justice to the enormous amount of work which has gone into the preparation of the manuscript. I can highly recommend this text to all students and researchers alike who want to start using stochastic differential equations in applications.

Contents:
PART I : Some General Techniques for Analysing Queueing and Related Models
1. Level-crossing analysis of queues, B. Doshi
2. Random allocation in a waiting room problem, J. Gani
3. Dynamic analysis of matrix Lindley process with replacement, Y. Masuda and U. Sumita
4. Reversibility and compound birth-death and migration processes, R.F. Serfozo
PART II : Some Queueing Models
5. Feedback retrial queueing systems, B.D. Choi and V.G. Kulkarni
6. A limit theorem on the output of GI/M/? queues, H. Kaspi and M. Rubinovitch
7. A tandem fluid network with Lévy input, O. Kella and W. Whitt
8. Second-order properties of single-stage queueing systems, L. Liyanage and J.G. Shanthikumar
9. On the relationship between stationary and Palm moments of backlog in the G/G/1 priority queue, M.A. Wortman and R.L. Disney
PART III: Approximations and Numerical Analysis
10. Inequalities concerning the waiting time in single-server queues: A survey, D.J. Daley, A. Ya. Kreinin and C.D. Trengove
11. An algorithm for finding characteristic roots of quasi-triangular Markov chains, C.M. Harris, W.G. Marchal and R.W. Tibbs
12. Approximating the distribution of the maximum queue length for M/M/s queues, W.P. McCormick and Y.S. Park
13. Numerical transient solution of finite Markovian queueing systems, J.K. Muppala and K.S. Trivedi
14. Approximations of performance characteristics in periodic Poisson queues, T. Rolski
PART IV : Control and Inference
15. Admission to a general stochastic congestion system: Comparison of individually and socially optimal policies, M. Bartroli and S. Stidham, Jr.
16. Sequential inference for single-server queues, I.V. Basawa and B.R. Bhat
17. Some results on inference for stationary processes and queueing systems, C.C. Heyde

This book, dedicated to N.U. Prabhu, is a collection of invited and refereed papers covering a variety of recent research in queueing theory. Several of the chapters illustrate new computational techniques and applications, others are reviews of previously scattered results together with further developments, and others present theoretical advances. Taken as a whole, the book certainly succeeds in giving an accurate reflection of the nature of a wide cross-section of cur-rent research, and should be of interest to both beginning and established research workers in this area.

Contents:
Introduction
PART I : Preliminaries
PART II : Discrete Theory
PART III: Continuous Theory
PART IV : Inequalities
PART V : Intensity-Governed Processes
PART VI : Diffusions

Over the last two decades coupling techniques have opened new vistas, sharpened insights, and provided incisive proofs in many areas of probability. This book provides the first systematic treatment of the subject. It is aimed at readers with a sophisticated understanding of probability. In keeping with the title, the style of the book is compact, however technicalities are fully sign posted and in some cases treated in detail. Perhaps in contrast to the implication of the title, the coverage is both broad and thorough. In some cases, for example, comparison techniques, it serves to show what is possible, easily, from the coupling method. In others, notably renewals and regeneration, it gives an essentially complete treatment, using coupling, of much of the classical theory. It is a reflection on both the beauty of the method and on the skills of the author that the book is so enjoyable to read.

Contents:
1. One-dimensional random walks
2. Two-dimensional random walks
3. The two-dimensional workload process
4. The N-dimensional random walk

In the last few years, several books have appeared on random walks. Most of them are mostly de-voted to weak and strong limit theorems. The present book deals with exact results like the characterization of the generating functions of different sequences of random variables and stochastic processes. The most important models are the reflecting and absorbing random walks and workload processes in the one- and two-dimensional cases. Some of the results are extended to the n-dimensional random walk.

Contents:
PART I : Preliminaries
1. Markovania
2. Passage probabilities
PART II : Analytic Theory
3. First entrance times
4. Last exit times
5. Dirichlet problem and Poisson equation
6. Decomposition theorems
7. Examples
PART III: Measure Theory
8. Functionals and potentials
9. Transformations
10. Extensions
PART IV : Applications
11. Structures
12. Boundary theory
13. Examples

"The book is a survey of work on passage times in stable Markov chains with a discrete state space and a continuous time parameter". The main language used is that from probabilistic potential theory. Besides giving a unifying approach to passage time problems, the author spends much time on discussing various applications to specific chains. As such, the reader gets a good overview of the potential applicability of certain techniques without getting lost in some of the more intricate measure theoretic problems encountered for general Markov processes. The latter theory is appropriately referenced throughout the book. The text can be used for a postgraduate course on the topic or isolated chapters may be brought in to complement a standard course on Markov chains. Students as well as researchers with an interest towards applications will find this a useful book.

Contents:
PART I : Critical Phenomena, Quantum Field Theory, Random Walks and Random Surfaces: Some Perspectives
1. General introduction
2. Phase transitions and critical points in classical spin systems: A brief survey
3. Scale transformation and scaling (continuum limits in lattice spin systems)
4. Construction of scaling limits: The renormalization group
5. Random walks as Euclidean field theory (EFT)
6. EFT as a gas of random walks with hard-core interactions
7. Random surface models
PART II : Random-Walk Models and Random-Walk Representations of Classical Lattice Spin Systems
8. Introduction
9. Random-walk models in the absence of magnetic field
10. Random-walk models in the presence of a magnetic field
11. Factorization and differentiation of the weights
12. Correlation inequalities: A survey of results
PART III: Consequences for Critical Phenomena and Quantum Field Theory
13. Background material
14. Inequalities for critical exponents
15. Continuum limits

This book provides a mathematical treatment of issues related to universality in statistical physics. Universality is the principle that "physical sys-tems are divided, according to their critical behaviour, into a relatively small number of classes". These classes are determined by the values of and relations among their critical exponents, which are the powers describing the divergence of quantities of physical interest at critical points in the parameter space. Information about critical exponents for many models has been obtained numerically and by the use of renormalization group arguments. Rigorous results about them are usually derived using correlation inequalities. Many of these inequalities, on the other hand, are based on random walk-type representations for the models of interest. In this work, three of the more important contributors to our understanding of this circle of ideas have provided us with a wealth of information in a clear and attractive form.

Contents:
1. Introduction
2. The biology of neural networks: A few features for the sake of non-biologists
3. The dynamics of neural networks: A stochastic approach
4. Hebbian models of associative memory
5. Temporal sequences of patterns
6. The problem of learning in neural networks
7. Learning dynamics in 'visible' neural networks
8. Solving the problem of credit assignment
9. Self-organization
10. Neurocomputation
11. Neurocomputers
12. A critical view of the modelling of neural networks

Readership: Graduate students, physicists and applied mathematicians, neurobiologists

Given the extraordinarily large numbers of neurons comprising the central nervous systems of higher organisms, a statistical approach in the sense of statistical physics to the theory of neural networks is very attractive. This book addresses both physicists and applied mathematicians, for whom an introduction to the biological background is provided, and neuro-biologists who want to go beyond the traditional descriptive approach of neurobiology, though they must be able to handle the mathematics. The author states that 'the book presents the status (sic) of the art in the field of neural networks that has been reached by the end of 1988', which makes one feel a little uneasy given the acknowledged pace of development in the field. Indeed, in the preface the author stresses that a single volume can nowadays only hope to serve as an introduction to what has become a very large subject. Nevertheless, he has made a creditable attempt to be comprehensive. The English is a little stilted in places, but the forty-five pages of references will provide a useful source.

Contents:
1. Introduction
2. An overview of database design
3. The entity-relationship model
4. The relational model
5. Object-oriented data models
6. Design principles
7. Integrity constraints and dependencies
8. Properties of relational schemas
9. Axiomatizations for dependencies
10. Algorithms for design problems
11. Mappings between ER-diagrams and relational database schemas
12. Schema transformations
13. Efficient algorithms for design problems
14. Use of example database in design
15. Dependency inference

Readership: Final year undergraduate and postgraduate students of database design, database design practitioners

The first chapter of the book points out some drawbacks of traditional methods for designing relational databases and states that the main goal of the book is to develop a theory of database design, 'De-sign-by-example', which overcomes them. This might make it sound as if the book is rather specialized, but in fact it does give a good introduction to relational database technology. Algorithms and methods for the use of example databases in database design are presented and the software tool 'Design-by-example' is currently under development, with release intended in 1993. Exercises and bibliographic notes are included at the end of each chapter. No prior knowledge of database systems of design is assumed and overall I found this a clear and well-written book and would recommend it.

Title	ALGORITHMS, ROUTINES, AND S FUNCTIONS FOR ROBUST STATISTICS. THE FORTRAN LIBRARY ROBETH WITH AN INTERFACE TO S-PLUS.
Author	A. Marazzi. With the collaboration of J. Joss and A. Randriamiharisoa.
Publisher	Pacific Grove, California: Wadsworth and Brooks/Cole, 1993, pp. xii + 436, US$59.95.

Reviewer:
Institute	Sevastopol Instrument Making Institute
Place	Sevastopol, Ukraine
Name	A.V. Tsukanov

Title	INTRODUCTION TO SHAPE OPTIMIZATION. Shape Sensitivity Analysis.
Author	J. Sokolowski and J.-P. Zolesio.
Publisher	Berlin: Springer-Verlag, 1992, pp. 250, DM.128.00.

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

Title	NUMERICAL METHODS FOR STOCHASTIC CONTROL PROBLEMS IN CONTINUOUS TIME.
Author	H.J. Kushner and P.G. Dupuis.
Publisher	New York: Springer-Verlag, 1992, pp. ix + 439, DM.98.00.

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

Reviewer:
Institute	Mill House, Hurst Green
Place	Brightlingsea, U.K.
Name	G.A. Barnard

Reviewer:
Institute	Stanford University
Place	Stanford, U.S.A.
Name	I. Olkin

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

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

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

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

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

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

Reviewer:
Institute	Statistics Canada
Place	Ottawa, Canada
Name	A.R. Gower

Reviewer:
Institute	Canadian Climate Centre
Place	Downsview, Canada
Name	F.W. Zwiers

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

Reviewer:
Institute	University of Manchester, Institute of Science and Technology
Place	Manchester, U.K.
Name	P.J. Laycock

Reviewer:
Institute	Queen's University
Place	Kingston, Canada
Name	J.D. Myles

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

Reviewer:
Institute	Katholieke Universiteit
Place	Leuven, Belgium
Name	B. Boone

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

Reviewer:
Institute	Pennsylvania State University
Place	University Park, U.S.A.
Name	B.G. Lindsey

Reviewer:
Institute	City University
Place	London, U.K.
Name	H.P Wynn

Reviewer:
Institute	University of Aarhus
Place	Aarhus, Denmark
Name	M. Sørensen

Reviewer:
Institute	University of Waterloo
Place	Waterloo, Canada
Name	J.F. Lawless

Reviewer:
Institute	University of Bonn
Place	Bonn, Germany
Name	M. Röckner

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

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

Reviewer:
Institute	Technical University of Vienna
Place	Vienna, Austria
Name	P. Révész

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

Reviewer:
Institute	University of Zürich
Place	Zürich, Switzerland
Name	E. Bolthausen

Reviewer:
Institute	BP Exploration
Place	Sunbury-on-Thames, U.K.
Name	D. Wolstenholme

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

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

Reviewer:
Institute	Mobil Research and Development Corporation
Place	Dallas, U.S.A.
Name	M.A. Maes