Short Book Reviews On-Line 1997

BONFERRONI-TYPE INEQUALITIES WITH APPLICATIONS.	J. Galambos. and I. Simonelli.
SAMPLING IN DIGITAL SIGNAL PROCESSING AND CONTROL.	A. Feuer and G.C. Goodwin.
ALGORITHMS FOR LINEAR-QUADRATIC OPTIMIZATION.	V. Sima.
MODIFIED LAGRANGIAN AND MONOTONE MAPS IN OPTIMIZATION.	E.G. Golshtein and N.V. Tretyakov. Translated by N.V. Tretyakov.
ERROR AND THE GROWTH OF EXPERIMENTAL KNOWLEDGE.	D.G. Mayo.
A PROBABILISTIC ANALYSIS OF THE SACCO AND VANZETTI EVIDENCE.	J.B. Kadane and D.A. Schum.
BAYESIAN METHODS AND ETHICS IN A CLINICAL TRIAL DESIGN.	J.B. Kadane (Ed.).
FRONTIERS OF ILLUSION: SCIENCE, TECHNOLOGY, AND THE POLITICS OF PROGRESS.	D. Sarewitz.
AGAINST THE GODS: THE REMARKABLE STORY OF RISK.	P.L. Bernstein.
ADVANCES IN BIOMETRY. 50 YEARS OF THE INTERNATIONAL BIOMETRIC SOCIETY.	P. Armitage and H.A. David (Eds.).
COMPUTATIONAL STATISTICS IN CLIMATOLOGY.	I. Polyak.
ELEMENTS OF PATTERN THEORY.	U. Grenander.
L'IMPLICATION STATISTIQUE. Nouvelle méthode exploratoire de données. Applications à la didactique.	R. Gras
OPERATIONAL SUBJECTIVE STATISTICAL METHODS: A Mathematical, Philosophical, and Historical Introduction.	F. Lad.
SIMULATION, 2nd edition.	S.M. Ross.
ELEMENTS OF SURVEY SAMPLING.	R. Singh and N. Singh Mangat.
HANDBOOK OF ITEM RESPONSE THEORY.	W.J. van der Linden and R.K. Hambleton (Eds.).
LATENT CLASS AND DISCRETE LATENT TRAIT MODELS.	T. Heinen.
MODERN MULTIDIMENSIONAL SCALING: THEORY AND APPLICATIONS.	I. Borg and P. Groenen.
TOOLS FOR STATISTICAL INFERENCE. Methods for the Exploration of Posterior Distributions and Likelihood Functions, 3rd edition.	M.A. Tanner.
PROBABILITY AND STATISTICAL INFERENCE.	R. Bartoszyñski and M. Niewiadomska-Bugaj.
PROBABILITY: SURVEY OF THE MATHEMATICAL THEORY, 2nd edition.	J.W. Lamperti.
ASPECTS OF STATISTICAL INFERENCE.	A.H. Welsh.
METHODS AND APPLICATIONS OF LINEAR MODELS. Regression and the Analysis of Variance.	R.R. Hocking.
LINEAR MODELS: A MEAN MODEL APPROACH.	B.K. Moser.
LINEAR AND NONLINEAR MODELS FOR THE ANALYSIS OF REPEATED MEASUREMENTS.	E.F. Vonesh and V.M. Chinchilla.
PRACTICAL DATA ANALYSIS FOR DESIGNED EXPERIMENTS.	B.S. Yandell.
INTRODUCTION TO THE DESIGN AND ANALYSIS OF EXPERIMENTS.	G.M. Clarke and R.E. Kempson.
INTRODUCTION TO TIME SERIES AND FORECASTING.	P.J. Brockwell and R.A. Davis.
TIME SERIES ANALYSIS. NONSTATIONARY AND NONINVERTIBLE DISTRIBUTION THEORY.	K. Tanaka.
PRACTICAL ANALYSIS OF EXTREME VALUES.	J. Beirlant, J.L. Teugels and P. Vynckier.
MARKOV PROCESSES FOR STOCHASTIC MODELING.	M. Kijima.
ASTROSTATISTICS.	G.J. Babu and E.D. Feigelson.
BAYES AND EMPIRICAL BAYES METHODS FOR DATA ANALYSIS.	B.P. Carlin and T.A. Louis.
BAYESIAN APPROACH TO INTERPRETING ARCHAEOLOGICAL DATA.	C.E. Buck, W.G. Cavanagh and C.D. Litton.
HIDDEN MARKOV AND OTHER MODELS FOR DISCRETE-VALUED TIME SERIES.	I.L. MacDonald and W. Zucchini.
MAXIMUM ENTROPY ECONOMETRICS: ROBUST ESTIMATION WITH LIMITED DATA.	A. Golan, G. Judge and D. Miller.
NEW APPROACHES TO MACROECONOMIC MODELING: EVOLUTIONARY STOCHASTIC DYNAMICS, MULTIPLE EQUILIBRIA, AND EXTERNALITIES AS FIELD EFFECTS.	M. Aoki.
ALGORTIHMES STOCHASTIQUES.	M. Duflo.
U-STATISTICS IN BANACH SPACES.	Y.V. Borovskikh.
FINANCIAL CALCULUS: AN INTRODUCTION TO DERIVATIVE PRICING.	M. Baxter and A. Rennie.
AN INTRODUCTION TO THE MATHEMATICS OF FINANCIAL DERIVATIVES.	S.N. Neftci.
PROBABILISTIC EXPERT SYSTEMS.	G. Shafer.
A FIRST COURSE IN OPTIMISATION THEORY.	R.K. Sundaram.
AN INTRODUCTION TO BAYESIAN NETWORKS.	F.V. Jensen.
L'ANALYSE DES DONNÉES ÉVOLUTIVES. Méthodes et Applications.	F. Dazy and J.-F. Le Barzic.
STATISTICAL INFERENCE. Based on the Likelihood.	A. Azzalini.
PUTTING CHANCE TO WORK. ... A LIFE IN STATISTICS.	N. Krishnankutty.
CLINICAL TRIALS: A METHODOLOGIC PERSPECTIVE.	S. Piantadosi.
STATISTICAL EVIDENCE. A Likelihood Paradigm.	R. Royall.
THE STATISTICAL APPROACH TO SOCIAL MEASUREMENT.	D.J. Bartholomew.
DEALING WITH RISK. Why the Public and the Experts Disagree on Environmental Issues.	H.W. Margolis.
CHAOS AND ORDER IN THE CAPITAL MARKETS. A NEW VIEW OF CYCLES, PRICES AND MARKET VOLATILITY, 2nd edition.	E.E. Peters.
ARCH MODELS AND FINANCIAL APPLICATIONS.	C. Gouriéroux.
DISCRETE MULTIVARIATE DISTRIBUTIONS.	N.L. Johnson, S. Kotz and N. Balakrishnan.
INITIATION AUX TRAITEMENTS STATISTIQUES. METHODES, METHODOLOGIE.	B. Escofier and C.J. Pagès.
STATISTICAL ASPECTS OF QUALITY CONTROL.	C. Derman and S.M. Ross.
STATISTICAL METHODS FOR INDUSTRIAL PROCESS CONTROL.	D. Drain.
SURVIVAL ANALYSIS: TECHNIQUES FOR CENSORED AND TRUNCATED DATA.	J.P. Klein and M.L. Moeschberger.
SURVIVAL ANALYSIS WITH LONG-TERM SURVIVORS.	R. Maller and X. Zhou.
PATTERN RECOGNITION USING NEURAL NETWORKS: Theory and Algorithms for Engineers and Scientists.	C.G. Looney.
CONSTRUCTION AND ASSESSMENT OF CLASSIFICATION RULES.	D.J. Hand.
METHODS FOR STATISTICAL DATA ANALYSIS OF MULTIVARIATE OBSERVATIONS, 2nd edition.	R. Gnanadesikan.
ESSENTIAL WAVELETS FOR STATISTICAL APPLICATIONS AND DATA ANALYSIS.	R.T. Ogden.
MULTIVARIATE MODELS AND DEPENDENCE CONCEPTS.	H. Joe.
MODERN REGRESSION METHODS.	T.P. Ryan.
INTRODUCTION TO ROBUST ESTIMATION AND HYPOTHESIS TESTING	R.R. Wilcox.
AN INTRODUCTION TO MEASURE AND PROBABILITY.	J.C. Taylor.
MATHEMATICAL THEORY OF RELIABILITY OF TIME DEPENDENT SYSTEMS WITH PRACTICAL APPLICATIONS.	I.N. Kovalenko, N.Y. Kuznetsov and P.A. Pegg.
RANDOM ITERATIVE MODELS.	M. Duflo.
A MODERN APPROACH TO PROBABILITY THEORY.	B. Fristedt and L. Gray.
STOCHASTIC PARTIAL DIFFERENTIAL EQUATIONS. A Modeling, White Noise Functional Approach.	H. Holden, B. Oksendal, J. Uboe and T. Zhang.
A WEAK CONVERGENCE APPROACH TO THE THEORY OF LARGE DEVIATIONS.	P. Dupuis and R.S. Ellis.
INTRODUCTION TO PRACTICAL LINEAR PROGRAMMING.	D.J. Pannell.
PLAIN ANSWERS TO COMPLEX QUESTIONS, 2nd edition.	R. Christensen.
THE STATISTICAL THEORY OF SHAPE.	C.G. Small.
STATISTICAL METHODS: A GEOMETRIC PRIMER.	D.J. Saville and G.R. Wood.
UNBIASED ESTIMATORS AND THEIR APPLICATIONS. VOLUME 2: MULTIVARIATE CASE.	V.G. Vionov and M.S. Nikulin.
INTRODUCTION TO CODING AND INFORMATION THEORY.	S. Roman.
METHODS OF MOMENTS AND SEMIPARAMETRIC ECONOMETRICS FOR LIMITED DEPENDENT VARIABLE MODELS.	M.-J. Lee.
PARAMETRIC STATISTICAL INFERENCE.	J.K. Lindsey.
ADVANCED STATISTICS: Volume I: Description of Populations.	S.J. Haberman.
ANALYZING AND MODELING RANK DATA.	J.I. Marden.
SMOOTHING METHODS IN STATISTICS.	J.S. Simonoff.
SMOOTHNESS PRIORS ANALYSIS OF TIME SERIES.	G. Kitagawa and W. Gersch.
ROBUST STATISTICS, DATA ANALYSIS, AND COMPUTER INTENSIVE METHODS. In Honor of Peter Huber's 60th Birthday.	H. Rieder (Ed.).
STATISTICAL TOOLS FOR NONLINEAR REGRESSION: A PRACTICAL GUIDE WITH S-PLUS EXAMPLES.	S. Huet, A. Bouvier, M.-A. Gruet and E. Jolivet.
FRAMES AND RESOLVABLE DESIGNS: USES, CONSTRUCTIONS AND EXISTENCE.	S. Furino, Y. Miao and J. Yin.
ADAPTIVE SAMPLING.	S.K. Thompson and G.A.F. Seber.
A COURSE IN LARGE SAMPLE THEORY.	T.S. Ferguson.
INTRODUCTION TO STATISTICAL TIME SERIES, 2nd edition.	W.A. Fuller.
DISCRETE GAMBLING AND STOCHASTIC GAMES.	A.P. Maitra and W.P.Studderth.
STOCHASTIC CALCULUS: A PRACTICAL INTRODUCTION.	R. Durrett.
MEASURE THEORY AND PROBABILITY, 2nd edition.	M. Adams and V. Guillemin.
MEASURES AND PROBABILITIES.	M. Simonnet.

Contents:
General introduction
1. The method of indicators
2. The method of polynomials
3. The geometric method
4. The linear programming method
5. Multivariate Bonferroni-type inequalities
6. Classical problems of probability and applications in combinatorics
7. Applications in number theory
8. Statistical applications
9. Extreme value theory
10. Miscellaneous topics

Readership: Mathematicians, probabilists, mathematical statisticians, number theorists

This is a book for the aficionado. The authors' aim is to present all the Bonferroni-type inequalities of which they are aware and to highlight their applications in combinatorics, statistics, ex-treme value theory and number theory. Their definition of Bonferroni-type inequalities includes a very wide range of inequalities that yield upper and lower bounds and are derived by extensions of the method of inclusion and exclusion. They give ways of generating such inequalities as well as methods of proof. The envisaged audience comprises mathematical scientists of various kinds. The level of mathematics, though rigorous, is not, however, inherently difficult; the stated prerequisites are "elementary probability theory and an interest in combinatorial arguments." Also, "all applied topics are fully introduced and those sections do not require any previous knowledge of the subjects." To benefit most from this scholarly presentation of the subject matter, the main requirement for many readers will be perseverance in following through sometimes quite lengthy arguments whose general direction is not always readily apparent.

Contents:
1. Fourier analysis
2. Sampling and reconstruction
3. Analysis of discrete time systems
4. Discrete-time models of continuous deterministic systems
5. Optimal linear estimation with finite impulse response filters
6. Optimal linear estimation with state space filters
7. Periodic and multirate filtering
8. Discrete time control
9. Sampled data control
10. Generalized sample-hold functions
11. Periodic control of linear time-invariant systems
12. Multirate control
13. Optimal control of periodic systems

Readership: Control theorists, others with interests in estimation and control of linear dynamical systems

This book might be seen as an antidote to the prevalent 'sample it and forget it' attitude of much control theory literature. The authors emphasize the fact that many of the systems which are analyzed as discrete time models acutally evolve in continuous time, sampled at discrete instants. The book casts Fourier analysis results in a form useful for analyzing such problems, and treats systems including sample and hold elements in a unified fashion. A range of problems is treated, from Kalman filtering and optimal control, through multirate and periodic problems. The book emphasizes a unity of approach to the various problems, and avoids rigorous derivations of most of the quoted results. This makes it most suitable as an interesting supplement to introductory treatments of the subject area.

Contents:
1. Linear-quadratic optimization problems
2. Newton algorithms
3. Schur and generalized Schur algorithms
4. Structure-preserving algorithms
APPENDIX A: Comparison of Riccati Solvers
APPENDIX B: Notation and Abbreviations

This text is intended to give theoretical, algorithmic and computational advice and encouragement in solving linear-quadratic optimization problems that arise in control engineering. For each numerical technique the author presents a mathematical description of the method and algorithm. He discusses the implementation of the algorithm along with its applicability and limitations. The algorithms are presented in a modified MATLAB notation. It is clear from the detailed recommendations about the computational choices that the author has coded and used the algorithms. In Appendix B he reviews his experiences in solving a set of problems using the EISPACK and LINPACK routines on an IBM PC 486. The problems range from small "real-world" that have been presented in the literature to randomly generated ones. This book is of value to anyone who has to, or who wishes to be prepared to, solve linear-quadratic control problems. The reader has to under-stand matrix algebra and have some experience of implementing a numerical algorithm.

Contents:
1. Introduction to convex analysis
2. Modified Lagrangian functions for convex programming problems
3. Dual methods
4. Monotone maps
5. Gradient-type methods and modification of a monotone map
6. Saddle gradient methods
7. Modified Lagrangian functions for smooth mathematical programming problems and related dual methods

This text covers the theory and, from a pure theoretical standpoint, the computational methods of modified Lagrangian functions. The term augmented Lagrangian is commonly used in the literature for a specific modified Lagrangian function used to solve a non-linear program. Modified Lagrangian functions are covered in only a few books of which the most well known is Constrained Optimization and Lagrange Multiplier Methods by D. Bertsekas (1978). This work is highly specialized and to fully appreciate it I would expect a reader to be a researcher in this area.

Contents:
1. Learning from error
2. Ducks, rabbits, and normal science: Recasting the Kuhn's-Eye view of Popper
3. The new experimentalism and the Bayesian way
4. Dunhem, Kuhn, and Bayes
5. Models of experimental inquiry
6. Severe tests and methodological underdetermination
7. The experimental basis from which to test hypotheses: Brownian motion
8. Severe tests and novel evidence
9. Hunting and snooping: Understanding the Neyman- Pearson predesignationist stance
10. Why you cannot be just a little bit Bayesian
11. Why Pearson rejected the Neyman-Pearson(behavioristic) philosophy and a note on objectivity in statistics
12. Error statistics and Peircean error correction
13. Towards an error-statistical philosophy of science

Readership: Statisticians interested in foundations of inference, philosophers of science

In recent years philosophers of science writing about statistical inference have tended to embrace a Bayesian approach. This book takes an opposing tack, with an extensive critique of Bayesian inference and a strong advocacy of a Neyman-Pearsonian frequentist approach. The book is clearly written, and al-though the criticisms of Bayesian statistics and the likelihood principle that are presented will not be new to statistical students of these debates, the framing of these questions within the current dialogue of the philosophy of science may well be. Some statistical readers may find the author's introduction of terminology an impedance, for example, "severity" for "power", "error statistics" for "frequency approach" or "sampling theory". Tests of hypotheses are emphasized, and problems posed for frequentist inference by conditional inference are not addressed. Glenn Shafer's idea of protocols is not cited, although it would have helped clear up some of the issues discussed.

Contents:
1. Different wine in an old bottle
2. A standpoint for our analysis of the Sacco and Vanzetti evidence
3. Chains of reasoning from a mass of evidence
4. Grading the probative force of the Sacco and Vanzetti evidence
5. Probabilistic analyses: Issues and methods
6. Probabilistic analyses: Judgements and stories
7. Probabilistic analyses of evidence in various disciplines
8. Final thoughts about Nicola Sacco and Bartolomeo Vanzetti
APPENDIX A: Wigmore Evidence Charts and Key Lists for the Trial and Post-Trial Evidence
APPENDIX B: Trial Witness List
APPENDIX C: Probabilistic Story Ingredients in Words

In countries where the criminal law rests on trial by jury, the jury is required to determine, on the basis of the evidence laid before it, whether the accused is guilty "beyond reasonable doubt", or not guilty. Sacco and Vanzetti were two anarchists convicted in 1920 of first degree murder. After many appeals, they were executed in 1927.
From p. 283: "Asked to render a verdict as far as Sacco and Venzetti are concerned, our vote, would be Vanzetti: innocent; Sacco: not proven".
The book is not easy going. The breadth of readership indicated means that statisticians will find many sections of the text obvious, and legal experts will find the same of many other sections, with little overlap between these sections. But a meeting of minds of the two professions is so much needed that reading it could be regarded as a duty.

Contents:
PART I : Major Issues
1. Introduction, by J.B. Kadane
2. Ethically optimizing clinical trials, by K.F. Schaffner
3. Admissibility of treatments, by N. Sedransk
4. Statistical issues in the analysis of data gathered in the new designs, by J.B.Kadane and T. Seidenfeld
PART II : Test Case: Verapamil/Nitroprusside
5. Introduction to the verapamil/nitroprusside study, by J.B. Kadane
6. The mechanics of conducting a clinical trial, by E.S. Heitmiller and T.J.J. Blanck
7. The verapamil/nitroprusside study: Comments on "The Mechanics of Conducting a Trial", by J.L. Coulehan
8. Computation aspects of the verapamil/nitroprusside study, by L.A. Galway
9. Being an expert, by T.J.J. Banck, T.J. Conahan, R.G. Merin, R.L. Pragner and J.J. Richter
10. Issues of statistical design, by N. Sedransk
11. Operational history and procedural feasibility, by J.B. Kadane
12. Verapamil versus nitroprusside: Results of the clinical trial I, by J.B. Kadane and N. Sedransk
13. Verapamil versus nitroprusside: Results of the clinical trial II, by E.S. Heitmiller, J.B. Kadane, N. Sedransk and T.J.J. Blanck
PART III: Other Issues
14. The law of clinical testing with human subjects: Legal implications of the new and existing methodologies, by D. Kairys
15. Commentary I on "The Law of Clinical Testing with Human Subjects", by D. Moore and A.J. Popp
16. Commentary II on "The Law of Clinical Testing with Human Subjects", by K.D. Katz
17. Author's response to Commentaries I and II, by D. Kairys
18. Whether to participate in a clinical trial: The patient's view, by L.J. Emrich and N. Sedransk
PART IV : Epilogue
19. Epilogue, by J.B. Kadane

This book discusses the issues involved in the design, conduct and analysis of a clinical trial. To this end the views of lawyers, medical doctors, patients, philosophers and statisticians are represented. A Bayesian framework within which a clinical trial may be carried out is described. This framework is then applied to a study comparing two drugs which may be used to control hypertension just after open-heart surgery. A novel aspect is the following. A panel of experts were utilized in order to determine whether a patient with certain risk factors should be randomized to one of the two drugs. If, in the experts' view, one of the treatments was clearly superior then the patient was assigned to that treatment. This book will contain much of interest to anyone who has an interest in the fascinating, controversial but undoubtedly important area of clinical trials.

Contents:
1. The end of the age of physics
2. The myth of infinite benefit
3. The myth of unfettered research
4. The myth of accountability
5. The myth of authoritativeness
6. The myth of the endless frontier
7. Pas de trois: Science, technology, and the marketplace
8. Science as a surrogate for social action
9. Toward a new mythology

At one level, this is the story of a personal conversion, a transformation from the world-view of a basic researcher in the earth sciences to that of a political aide specializing in science and technology issues at the U.S. Congress. With empathetic fascination, the reader watches as scientist Sarewitz recognizes and confronts the five "myths" that have shaped his own view of the role of science and technology in society, and also the view of most researchers.
At a deeper level, however, this is an appeal to scientists and engineers to follow Sarewitz along this road to Damascus C to abandon the obviously inadequate paradigms that have sustained public funding of research and development in the post-war era. Since more science does not automatically confer to better society, how do we ensure greater societal benefit from whatever research and development is done?
Recognizing that even asking such a question is anathema to many researchers, Sarewitz convincingly demonstrates that these scientists are blinkered by a world-view that does not stand up to the level of rigorous scrutiny these same scientists would demand in their own specialized sub-discipline. Unfortunately, in both his description of the problem and his tentative suggestions toward a solution, the author suffers from a similar tunnel vision.
The trouble lies in the book's obsessive focus on the societal management of science and technology solely in the United States. Of course, this is the world which Sarewitz knows first-hand, but it is hardly the solar system, much less the galaxy. It is surely questionable philosophy (and lousy statistics) to shape a new value system for something as universal as science and technology based on a sample size of one, even if that one dominates global research and development.
To give just one example of how this tunnel vision diminishes the authority of the book, Sarewitz easily demonstrates that the priorities for research and development in the industrialized world (read, the U.S.) seldom match the needs of the Third World. He then speculates about possible institutional responses to this gap, in apparent ignorance about the leading example, the International Development Centre in Ottawa, still going strong after two decades.
The obsessive American focus also explains some of the author's more bizarre and exasperating side-trips, including the most perverted account of the origins and lessons of the Gulf War outside of Noam Chomsky. As Sarewitz himself points out, it is a bit thick to expect the rest of the world to take lessons in informed decision-making from a system which cannot even legislate a seat-belt law.
With these caveats, Frontiers of Illusion is nonetheless an important book because it is written by a researcher who has managed to break free of the shackles with which so many scientists voluntarily (and unconsciously) blinker their world-view.

Contents:
1200: Beginnings
1. The winds of the Greeks and the role of the dice
2. Easy as I, II, III
1200-1700: A Thousand Outstanding Facts
3. The renaissance gambler
4. The French connection
5. The remarkable notions of the remarkable notions man
1700-1900: Measurement Unlimited
6. Considering the nature of man
7. The search for moral certainty
8. The supreme law of unreason
9. The man with the sprained brain
10. Peapods and perils
11. The fabric of felicity
1900-1960: Clouds of Vagueness and the Demand for Precision
12. The measure of our ignorance
13. The radically distinct notion
14. The man who counted everything except calories
15. The strange case of the anonymous stockbroker
Degrees of Belief: Exploring Uncertainty
16. The failure of invariance
17. The theory police
18. The fantastic system of side bets
19. Awaiting the wildness

Readership: General readership: statisticians or lay people interested in the history of the concepts of probability and risk

This book is aimed at the popular market and does not go into depth or technical detail. It falls naturally into two parts: the first provides a history of probability, and the second an informal description of early basic risk management strategies in financial markets. If you have read any of the standard works on the history of probability, such as Hacking's various books on the subject, then there will be very little to surprise you in the first part of the book. The emphasis on financial markets in the second part be-trays a very specialized view of the word 'risk' (al-though this is not surprising since the author is a professional investor and has written other books on economics and finance) - and the author does acknowledge that 'risk' impinges on all walks of life. A few minor errors grated on the present reviewer (I'll let you find them yourself), but there were not enough to detract from a general good impression of the book.

Contents:
1. The roads travelled: Our 50-year journey, by L. Billard
2. Chester Bliss and the International Biometric Society, by A.T. James
3. The International Biometric Society and Biometrics: Contributions to experimental design, by R.A. Bradley and R.L. Anderson
4. The teaching of biometry, by P. Dagnelie
5. Linear models 1945-1995, by S.R. Searle and C.E. McCulloch
6. The evolution of categorical data modelling: A biometric perspective, by P.B. Imrey, G.G. Koch and J.S. Preisser
7. Empirical Bayes and hierarchical Bayes procedures in simultaneous estimation of parameters, by C.R. Rao
8. Computer-intensive statistical methods, by B. Efron and R. Tibshirani
9. Multivariate and multidimensional analysis, by J.C. Gower
10. Survival analysis, by P.K. Andersen and N. Keiding
11. Statistical computing, by J.A. Nelder
12. Statistical ecology, environmental statistics and risk assessment, by G.P. Patil
13. Statistics for agriculture and forestry, by G.H. Freeman and J. Riley
14. Statistical genetics, by E.A. Thompson
15. Statistics in epidemiology: The case-control study, by N.E. Breslow
16. Biometric advances in infectious disease epidemiology, by K. Dietz
17. The design and analysis of longitudinal studies: A historical perspective, by J.H. Ware and K.-Y. Liang
18. Spatial analysis in biometry, by P.J. Diggle
19. A review of image analysis in biometry, by C.A. Glasbey and M. Berman
20. Clinical trials: A statistician's perspective, by S.J. Pocock
21. Statistical analysis of toxicological experiments on carcinogenicity, mutagenicity, and developmental toxicity, by D. Krewski, B.G. Leroux and Y. Zhu

Readership: Biostatisticians, statisticians, the rare biological or medical scientist

This collection of review articles is a celebration of the International Biometric Society and its journal, Biometrics. The chapter titles and list of world-class authors promise a superb anniversary present as soon as the package is opened. The editors' goal was to cover topics of persisting concern to the International Biometric Society, both methodological and applied. Contributors were asked to portray the current state of their specialities and to trace developments of the past fifty years. According to the editors, "Authors were allowed to interpret the broad guidelines in their own way: as befits a society devoted to the study of biological variation, there were marked differences of approach." Nonetheless, the book is a sophisticated, cohesive overview of the field; thanks, one surmises, to the historical insight and influence of the editors.

Contents:
1. Digital filters
2. Averaging and simple models
3. Random processes and fields
4. Variability of ARMA processes
5. Multivariate AR processes
6. Historical records
7. The GCM validation
8. Second moments of rain

Readership: Climatologists, statistical software developers, statistical modellers

As the title indicates, the focus of this book is very much on computational statistics. In particular, the first five chapters concentrate on computational schemes for respectively: the normal equations; different schemes for averaging data, both temporally and spatially; correlation functions and spectra; univariate ARMA models; multivariate AR models. The author's stated goal is to present numerical algorithms in a form immediately useful for software development. Chapters 3, 4 and 5 each have relatively brief final sections where climatological examples are discussed, but it is not until two-thirds of the way through the book, in Chapter 6, that the author starts to discuss climatological issues at any length. For me, these chapters were the most interesting, although the role of statistics in detecting climate change is rather disappointingly relegated to two and a bit pages where it is emphasized that available data are not sufficiently precise to address the issues.

Contents:
1. The search for order
2. Open patterns
3. Closed patterns
4. Analyzing patterns
5. Analysis of open patterns
6. Analysis of closed patterns
7. Computer experiments with patterns
8. Computing open patterns
9. Computing closed patterns
10. More pattern experiments

This is the most accessible introduction to Grenander's theory of patterns (not to be confused with pattern recognition) and it would be ideal as preliminary reading for his 1993 monograph on General Pattern Theory, [Short Book Reviews, Vol. 14, p.43]. The latter has the full theory whereas this volume discusses some every-day examples to give a catalogue of patterns (the first three chapters), an introduction to the theory (the next three), and some computer experiments with MATLAB code. The final chapter contains a set of experiments for the assiduous reader to try out. There is a fair amount of mathematical notation in the middle part but very little mathematical knowledge is assumed.
The scope of examples is amazing: there are colour plates of paintings of battles, jigsaws, textile patterns, shadows of hands, a potato and a fairy ring, and line figures of salt crystal, leaves, mitochondria, a Pap smear, root systems and sea ice. Grenander's work deserves a wide audience, and this book deserves to be read widely.

Avec la colloboration de S. Almouloud, M. Bailleul, A. Larher, M. Polo, H. Ratsimba-Rajon, A. Totohasina. Préface par E. Diday.

Sommaire
Partie I : La Théorie
1. La théorie de l'implication statistique
Partie II: Applications de la Méthode
2. Analyse de comportements d'élèves dans de courtes démonstrations d'algèbre et de géométrie
3. Analyse de comportements d'élèves dans des démonstrations proposées par le logiciel D.E.F.I.
4. Mise en évidence et validation de conceptions d'étudiants sur le notion de probabilité conditonnelle
5. Analyse des phénomènes d'obsolence et de processus d'ostension
6. L'analyse implicative comme outil d'investigation des représentations de l'enseignement des mathématiques
7. Apports de la notion de contribution à l'analyse d'un questionnaire sur le répère cartésien

La méthodologie dans ce livre a été développée à l'Université de Rennes dans des thèses que l'auteur principal a dirigées. Le livre consiste de deux parties: la théorie et les applications en didactique. Dans la partie théorique, le concept fondamental est l'indice d'implication entre variables (binaires). Cette définition est construite sur des bases probabilistes et elle quantifie la plausibilité de la règle, "si a alors b". Dans l'étude des propriétés il y a évidemment des relations avec la logique mathématique et les concepts dans la statistique habituelle (par example corrélation, dépendance,...). Une grande différence est la dissymétrie dans la définition. Dans la seconde partie du livre, les auteurs présentent des applications de la méthode d'analyse implicative en donnant des examples sur le plan didactique. Les illustrations concernent l'analyse des données sur des situations d'enseignement de la mathématique aux élèves et aux étudiants. En conclusion, ce livre m'a paru très remarquable et aura l'air très inhabituel pour la majorité des statisticiens. Mais les chercheurs en didactique trouveront ici une richesse d'idées pour des méthodes alternatives en analyse des données.

Contents:
1. Philosophical and historical introduction
2. Quantities, prevision, and coherency
3. Coherent statistical inference
4. Related forms for asserting uncertain knowledge
5. Distribution functions
6. Proper scoring rules
7. The multivariate normal distribution and its mixtures
8. Sequential forecasting based on linear conditional prevision structures: theory and practice of linear regression
9. The direction of statistical research

Readership: Four distinct types of reader: (1) undergraduate statistics students, (2) advanced undergraduate and even graduate students, (3) more advanced mathematical and statistical researchers and prospective teachers of statistics, (4) the generally educated adult who is concerned with questions regarding the role of personal beliefs and values in the conduct of science

This is an advanced introduction to the philosophy and theory of the subjective approach to probability and statistical inference. The ideas are set in a historical context and sufficient references are given that one can pursue both the philosophy and history at greater depth if one wishes. Although advanced, the author has an attractive and easy style which makes the book a pleasure to read. In particular, although he does not shy from mathematical formalism, he does not adopt unnecessarily sophisticated mathematics. He assumes that the reader will be supported by a teacher (for example, p.99. If you are completely unfamiliar with linear programming, your teacher will review for you basic ideas ...'), but again, the relaxed style means this will seldom cause problems. Support material for teachers using the text can be obtained from the author by e-mail. The book is liberally illustrated with small descriptive examples as well as extensive problems, which the author clearly regards as part of the learning experience, and not an optional extra. The book is not a practical primer in Bayesian data analysis but will give a very solid theoretical grounding for anyone undertaking such analysis. Whether one is Bayesian or not, the book is well worth close study. It is the sort of book which will repay re-reading.
The author says he hopes that 'this text will be found intriguing to university-level teachers of mathematical probability and statistics'. He certainly succeeded in my case. I think most of us can learn from this book: both about the substantive content for which he presents a fascinating and convincing case, and, more mundanely, about how to present sophisticated and advanced ideas in an attractive, accessible, and exciting way.

Contents:
1. Introduction
2. Elements of probability
3. Random numbers
4. Generating discrete random variables
5. Generating continuous random variables
6. The discrete event simulation approach
7. Statistical analysis of simulated data
8. Variance reduction techniques
9. Statistical validation techniques
10. Markov chain Monte Carlo methods
11. Some additional topics

This book provides a well-written introduction to the important topic of simulation. The book begins at a level which is not technical but still manages to cover a wide-range of topics and techniques though it is not, and does not profess to be, a comprehensive text. Each chapter ends with references and a set of exercises. The chapter on Markov chain Monte Carlo is a particularly useful addition.

Contents:
1. Collection of survey data
2. Elementary concepts
3. Simple random sampling
4. Sampling with varying probabilities
5. Stratified sampling
6. Systematic sampling
7. Ratio and product methods of estimation
8. Regression method of estimation
9. Two-phase sampling
10. Cluster sampling
11. Multistage sampling
12. Sampling from mobile populations
13. Nonresponse errors

As the title suggests, this is an elementary text which could be used as an introduction to sample survey methodology or as a reference text, for easy location of the appropriate formulae for the estimators and their variances in a wide variety of sampling situations. The usual compendium of sampling structures is developed, ranging from simple random sampling with replacement, through systematic and stratified sampling, to cluster sampling and multistage sampling. Each chapter discusses the structure of the problem, gives the notation and formulae for the estimators of means, totals or proportions and the associated formulae for their variances and the estimators of these variances. Each set of formulae is clearly and conveniently dis-played in a box, but no theoretical developments are included. Every situation is illustrated by at least one example and there are plenty of exercises at the end of the sections. Many of the examples reflect situations met by the authors in their experiences at the Punjab Agricultural University. The last two chapters cover a variety of capture-recapture techniques and randomized response methods. In the final section, describing randomized response techniques, the authors include several illustrative examples of extensions of these ideas produced by them in recent publications. The text provides a clear, concise and practical discussion of all the basic techniques.

Contents:
1. Item response theory: Brief history, common models, and extensions, by W.J. van der Linden and R.K. Hambleton
PART I : Models for Items with Polytomous Response Formats
2. The nominal categories model, by R.D. Bock
3. A response model for multiple-choice items, by D. Thissen and L. Steinberg
4. The rating scale model, by E.B. Andersen
5. Graded response model, by F. Samejima
6. The partial credit model, by G.N. Masters and B.D. Wright
7. A steps model to analyse partial credit, by N.D. Verhelst, C.A.W. Glas, and H.H. de Vries
8. Sequential models for ordered responses, by G. Tutz
9. A generalised partial credit model, by E. Muraki
PART II : Models for Response Time or Multiple Attempts on Items
10. A logistic model for time-limit tests, by N.D. Verhelst, H.H.F.M. Verstralen, and M.G.H. Jansen
11. Model for speed and time-limit tests, by E.E. Roskam
12. Multiple-attempt, single-item response models, by J.A. Spray
PART III: Models for Multiple Abilities or Cognitive Components
13. Unidimensional linear logistic Rasch models, by G.H. Fischer
14. Response models with manifest predictors, by A.A. Zwinderman
15. Normal-ogive multidimensional model, by R.P. McDonald
16. A linear logistic multidimensional model for dichotomous item response data, by M.D. Reckase
17. Loglinear multidimensional item response model for polytomously scored items, by H. Kelderman
18. Multicomponent response models, by S.E. Embretson
19. Multidimensional linear logistic models for change, by G.H. Fischer and E. Seliger
PART IV : Nonparametric Models
20. Nonparametric models for dichotomous responses, by R.J. Mokken
21. Nonparametric models for polytomous responses, by I.W. Molenaar
22. A functional approach to modelling test data, by J.O. Ramsay
PART V : Models for Nonmonotone Items
23. A hyperbolic cosine IRT model for unfolding direct responses of persons to items, by D. Andrich
24. PARELLA: An IRT model for parallelogram analysis, by H. Hoijtink
PART VI : Models with Special Assumptions about the Response Process
25. Multiple group IRT, by R.D. Bock and M.F. Zimowski
26. Logistic mixture models, by J. Rost
27. Models for locally dependent responses: Conjunctive item response theory, by R.J. Jannarone
28. Mismatch models for test formats that permit partial information to be shown, by T.P. Hutchinson

Readership: Practitioners and researchers in the behaviourial sciences, teachers of item response theory

Classical item response theory assumes binary responses and a univariate ability continuum. During the 1980s these basic models were extended to allow different types of response variable (nominal, graded, partial-credit, and so on) and more sophisticated representations of the 'ability' continuum (multivariate, replace ability by attitude, etc.). The aim of this edited collection of specially written chapters is to bring together these recent developments. The book has been very carefully edited. Almost all chapters have the same format: an introduction, a presentation of the model, details of parameter estimation, how to measure goodness-of-fit, an empirical example, and discussion. This means that the chapters can be read independently. It opens with a short historical chapter, surveying earlier work and setting the scene for the later chapters, and also provides brief introductions at the start of each part of the book. This leads to an effective integration. Overall the book provides a comprehensive overview and sourcebook to models of these kinds.

Contents:
1. Introduction
2. Log-linear models and latent class analysis
3. Latent class measurement models
4. Latent trait models
5. Discretized latent trait models
6. Estimation in latent trait models

Readership: Social scientists and psychologists concerned with measurement problems and the analysis of discrete data

The series editor describes latent structure models as `the most important, and the most interesting, contribution of the social behaviourial sciences to data analysis', mentioning, amongst other such models, Bayes methods, imputation methods, hidden Markov models, and state-space models, but he could also have added neural networks. This book presents a unified view of two types of latent-structure model: latent-class and latent-trait models. These are latent-structure models in which the manifest variables are discrete though they may have any level of measurement. (This means that the book excludes structural-equation models.) Latent-class models assume that the population has a finite mixture distribution, with the manifest variable being independent within each class. Latent-trait models assume a continuous latent variable. The two types of models are related to log-linear models, and the author assumes some knowledge of these. The book provides a useful introduction to the ideas. The level of mathematical sophistication required is not great. Interested readers might also examine the excellent recent book by Bartholomew [The Statistical Approach to Social Measurement, Academic Press, 1996. To be reviewed in Short Book Reviews].

Contents:
PART I : Fundamentals of MDS
1. The four purposes of multidimensional scaling
2. Constructing MDS representations
3. MDS models and measures of fit
4. Three applications of MDS
5. MDS and facet theory
6. How to obtain proximities
PART II : MDS Models and Solving MDS Problems
7. Matrix algebra for MDS
8. A majorization algorithm for solving MDS
9. Metric and non-metric MDS
10. Confirmatory MDS
11. MDS fit measures, their relations, and some algorithms
12. Classical scaling
13. Special solutions, degeneracies and local minima
PART III: Unfolding
14. Unfolding
15. Special unfolding models
PART IV : MDS Geometry as a Substantive Model
16. MDS as a psychological model
17. Scalar products and Euclidean distances
18. Euclidean embeddings
PART V : MDS and Related Methods
19. Procrustes procedures
20. Three-way Procrustean models
21. Three-way MDS models
22. Methods related to MDS
APPENDIX A: Computer Programs for MDS
APPENDIX B: Notation

This book is concerned with all those techniques that represent similarity, dissimilarity, or general proximity data as points in multidimensional space. Such techniques had their origins and flowering in psychological and psychometric studies, and these substantive links form a strong thread throughout the volume. The coverage is extremely comprehensive, ranging through descriptive, data analytic, mathematical, computational and substantive aspects. The treatment is modern and up-to-date, dealing particularly well with the many recent algorithmic and computational advances. An appendix describes the main computer programs currently available for doing multidimensional scaling and shows how simple multi- dimensional scaling analyses can be effected in each program.
The structure of the book makes it suitable for beginner and expert alike: the former can get a working grasp of the fundamentals from the first six chapters, while the latter should enjoy much of the detail in the later chapters. The last chapter dealing with links between multi-dimensional scaling and other multivariable methods is perhaps the least successful one, but overall this is a very useful book.

Contents:
1. Introduction
2. Normal approximations to likelihoods and to posteriors
3. Non-normal approximations to likelihoods and posteriors
4. The EM algorithm
5. The data augmentation algorithm
6. Markov chain Monte Carlo: The Gibbs sampler and the Metropolis algorithm

Readership: Applied statisticians, graduate students and statisticians interested in modern computer-oriented techniques

This is the third edition of a very useful book. The author has added new examples and discussion of much recent work, particularly on Markov Chain Monte Carlo sampling; this chapter alone has over fifty pages. The subtitle of the book is "Methods for the exploration of posterior distributions and likelihood functions". In the early chapters there is a very useful review of the most important statistical distributions, including the multivariate normal and Wishart distributions, and also introductions to likelihood and Bayesian theory, illustrated with excellent examples. Standard Monte Carlo methods are also described. The last three chapters make up the bulk of the book, and contain an array of computational algorithms with the purpose of the subtitle in mind. They are covered clearly and thoroughly, and important questions such as convergence rates and when the methods might not work are not neglected. Complicated mathematical details are omitted, but references are given. The algorithms are presented in a step-by-step style which is valuable for the practical user. In these chapters especially, many examples are given involving real data sets, and the techniques are illustrated with many useful figures. There are exercises at the end of every chapter. As the author emphasizes, the book is directed largely to Bayesians and to statisticians basing inference on the likeli-hood, and this compendium of techniques will help to make Bayesian solutions more readily available. How-ever, the last section shows how the algorithms may be used in a classical situation, the two-way contingency table. The book will interest all statisticians wanting to become familiar with modern computational techiques.

Contents:
1. Experiments, sample spaces, and events
2. Probability
3. Combinatorial probability
4. Conditional probability: Independence
5. Markov chains
6. Random variables: Univariate case
7. Random variables: Multivariate case
8. Expectation
9. Some probability models
10. Limit theorems
11. Outline of inferential statistics
12. Estimation
13. Testing statistical hypotheses
14. Discrimination
15. Linear models
16. Rank models
17. Analysis of categorical data

This book is very different from the usual textbook with definitions, theorems, proofs and examples. True, all these elements can be found in this book as well, but the style is surprisingly different. Concepts and methods are illustrated by examples that actually are the backbone of the book. Teachers search-ing for inspiring examples of stochastic thinking will easily find something to their taste. Methods are shows "why" they work, and not only "how". To help the more technically adventurous reader some chapters and/or sections are marked with asterisks. This is a refreshing book that can be strongly recommended, even for self study, i.e. by the great wealth of inspiring problems at the end of each section.

Contents:
1. Foundations
2. Laws of large numbers and random series
3. Limiting distributions and the central limit problem
4. The Brownian motion process
APPENDIX: Essentials of Measure Theory

This is quite a substantial revision of the 1966 first edition. The general approach appears similar, with much careful discussion and motivation of concepts, so that, as in the first edition, it seems less to resemble the familiar `theorem proof' style. Exercises (over a hundred) are distributed in the main text. Some analysis and elementary probability are prerequisites; basic concepts from measure theory, needed to appreciate much of the material, are treated in an appendix new to this edition. A good range of material is covered for a relatively short volume; for instance Chapters 2 and 3 include discussions respectively of the law of the iterated logarithm and of stable and infinitely divisible distributions. The reviewer feels this book succeeds well in giving some taste of a wide variety of probability theory.

Contents:
1. Statistical models
2. Bayesian, fiducial and likelihood inference
3. Frequentist inference
4. Large sample theory
5. Robust inference
6. Randomization and resampling
7. Principles of inference

This is a book on statistical inference which has at the same time a reasonable mathematical level and the right attention to data analysis which makes the concepts less abstract. This approach makes it different from most other books in the field. Throughout a critical discussion is given on the different approaches to statistical inference (Bayesian, fiducial, likelihood, frequentist). In addition to the classical topics, the author also included: robustness, randomization, finite population inference, bootstrap, smoothing, etc. These things make the book very interesting since they open a window to the more recent developments. The book is self-contained, with many exercises, suggestions for further reading, a mathematical appendix and a long bibliography.

Contents:
PART I : Introduction to Basic Theory
1. Introduction to linear models
2. The distribution of linear and quadratic forms
3. Estimation and inference in simple linear models
4. Simultaneous inference: Tests and confidence intervals
PART II : Regression Models
5. Regression on functions of one variable
6. Transforming the data and miscellaneous topics
7. Regression on the functions of several variables
8. Collinearity in multiple linear regression
9. Influential observations in multiple linear regression
10. Polynomial models and models with qualitative predictors
11. Related topics
PART III: Analysis of Variance Models
12. Fixed effects models: I. Single-factor classification means
13. Fixed effects models: II. Two-way cross- classification
14. Fixed effects models: III. Nested factors and general structure
15. Fixed effects models: I. The AOV with balanced data
16. Fixed effects models: II. The AVE method
17. Fixed effects models: III. Unbalanced data

Readership: Statistics specialists, for whom this is a textbook and a research source; statistics users, for whom this is a reference manual

Over the years, R. Hocking has written a series of regression articles, all readable and useful, and this book represents the logical next step. As the chapter headings indicate, this is a classical presentation of regression and analysis of variance models and would be a very appropriate text for a two-semester foundation presentation for regression users. Some useful topics lie concealed in the chapter headings. For example, Chapter 9 contains material on principal components and robust methods. Appendix D contains fourteen sets of data which can also be downloaded from the Wiley ftp server. The writing style is, as we have grown to expect, excellent, and this is a book that is easy to use and a pleasure to read. Recommended!

Contents:
1. Linear algebra and related introductory topics
2. Multivariate normal distribution
3. Distribution of quadratic forms
4. Complete, balanced factorial experiments
5. Least-squares regression
6. Maximum likelihood estimation and related topics
7. Unbalanced designs and missing data
8. Balanced incomplete block designs
9. Less than full rank models
10. The general mixed model

This is a very solid, comprehensive and still rather compact introduction of linear models. A few relatively simple ideas and tools from linear algebra and theory of quadratic forms of normally distributed variables are consistently used through the entire text. The text is well structured, maybe slightly overloaded with lengthy formulae, there being no obvious way to avoid the latter. I would recommend the book as the very competitive choice of the basic text-book for graduate classes on the theory of linear models.

Contents:
1. Preliminaries
2. The multivariate analysis of variance (MANOVA) model
3. ANOVA and MANOVA for repeated measures
4. Crossover experiments
5. The generalized multivariate analysis of variance(GMANOVA)
6. Linear mixed-effects models for repeated measurements
7. Nonlinear regression models for repeated measurements
8. Generalized nonlinear mixed-effects models
9. A summary of selected methods of estimation

Readership: Scientists with repeated measurements data to analyze, and/or who teach the subject

"Repeated measurements" describes data to which an experimental unit contributes multiple times. Often, but not always, these multiple contributions are observations at successive times. They could also be related, for example, to dose levels of a drug. This book talks about various ways such data can be analyzed. The authors' experiences, and examples, lean towards applications in health and life sciences and in epidemiology and biomedical research, but applications in other fields are many. This is a complex and expanding area and the authors have done an excellent job of getting the reader into it. Good matrix skills are essential for the reader; the review in Chapter 1 is potent but brief. The heavy computing requirements for the methods discussed are helped by the provision of a disk of "MIXLIN 3.1". This enables use of OLS, ordinary least squares, EGLS, estimated generalized least squares, IRGLS, iteratively re-weighted generalized least squares, AMLE, approximate maximum likelihood, provided the user has "access to SAS Base soft-ware and also to the SAS-Macro and SAS-IML facilities" (p.461). The bibliography lists about three hundred and seventy-five references. The author index even shows bibliography mentions; this is sensible for second, etc. authors and is, harmless, overkill for the first authors! Also harmless, but amusing, is the back-cover claim of "over 1300 display equations." Is this a boast, or a complaint?

Contents:
PART A: Placing Data in Context
1. Practical data analysis
2. Collaboration in science
3. Experimental design
PART B: Working with Groups of Data
4. Group summaries
5. Comparing several means
6. Multiple comparisons of means
PART C: Sorting Out Effects with Data
7. Factorial designs
8. Balanced experiments
9. Model selection
PART D: Dealing with Imbalance
10. Unbalanced experiments
11. Missing cells
12. Linear models inference
PART E: Questioning Assumptions
13. Residual plots
14. Comparisons with unequal variance
15. Getting free from assumptions
PART F: Regressing with Factors
16. Ordered groups
17. Parallel lines
18. Multiple responses
PART G: Deciding on Fixed or Random Effects
19. Models with random effects
20. General random models
21. Mixed effects models
PART H: Nesting Experimental Units
22. Nested designs
23. Split plot designs
24. General nested designs
PART I: Repeating measures on subjects
25. Repeated measures as split plots
26. Adjustments for correlation
27. Cross-over designs

We are presented with a very careful and detailed exposition of the analyses of normally distributed data obtained from some well-known experimental designs. The emphasis is very much on the analysis of data rather than experimental design. The author encourages good communication between the scientist and statistical consultant and the book successfully deals with the concerns of both parties.
There are a number of useful features. The mathematical content is clear and at an appropriate level, and a number of important topics are included, for example, model selection, analysis of residuals and transformations. The applications are realistic and sufficient numerical detail is given to illustrate the analyses.
Although some common designs (for example, Latin squares and balanced incomplete block designs) are only mentioned briefly, the inclusion of other more advanced topics (for example, restricted maximum likelihood estimation (REML)) transforms the book from an elementary text to one which would form a useful addition to the shelves of any scientist or applied statistician.

Contents:
1. Collecting data by experiments
2. Basic statistical methods: The normal distribution
3. Principles of experimental design
4. The analysis of data from orthogonal designs
5. Factorial experiments
6. Experiments with many factors: Confounding and fractional replication
7. Confounding main effects: Split-plot designs
8. Industrial experimentation
9. Response surfaces and mixture designs
10. The analysis of covariance
11. Balanced incomplete blocks and general non- orthogonal block designs
12. More advanced designs
13. Random effects models: Variance components
14. Computer output using SAS

This book covers the analysis of normally distributed data arising from standard experimental designs and is traditional in its approach. Practical applications are an important feature and numerous worked numerical examples are used effectively to illustrate the techniques described. The mathematical level is straightforward and would appeal to the scientist wanting to gain an understanding of the main concepts rather than to the theoretical statistician.
Much effort is put into explaining how to do calculations 'by hand'. For example, there is a chapter devoted to analysis of covariance and advice on how to adapt the standard analyses for missing observations. In an age when most calculations are performed using computing packages, this gives the book a slightly old-fashioned feel although there is a chapter devoted to output from SAS.
Useful features include an introduction to Taguchi methods and discussion of mixture designs and non-orthogonal block designs.

Contents:
1. Introduction
2. Stationary processes
3. ARMA models
4. Spectral analysis
5. Modelling and forecasting with ARMA processes
6. Nonstationary and seasonal time series models
7. Multivariate time series
8. State-space models
9. Forecasting techniques
10. Further topics

Readership: Undergraduate students in mathematics, statistics, engineering, economics, natural sciences and social sciences; anybody who wants a first course in time series analysis

Following the great success with their book Time Series: Theory and Methods first published in 1987 [Short Book Reviews, Vol.7. p.45], and revised in 1991, Brockwell and Davis have now written a more elementary text aiming 'at the reader who wishes to gain a working knowledge of time series and forecasting methods as applied in economics, engineering and the natural and social sciences'. They have done so by giving careful explanations of basic ideas and thoughtful illustrations of techniques. The interested readers are referred to their time series theory and methods for proofs of deeper theoretical results. The emphasis is on hands-on experience and the friendly software which accompanies the book serves the purpose admirably. There is now no excuse whatever for any self-respecting statistician (or even scientist) not knowing the ABC of time series analysis. In fact, the coverage of the book goes beyond the ABC by including topics of current research interests such as non-linear time series (including chaos), volatility, long memory models, continuous time modelling and so on. It is natural that there are a few typographical errors in this first edition of the book. As I expect the book to enjoy longevity, these errors will be short-lived. The authors should be congratulated for making the subject accessible and fun to learn. The book is a pleasure to read and highly recommended. I regard it as the best introductory text in town.

Contents:
1. Motivating examples
2. Stochastic calculus in mean square
3. Functional central limit theorems
4. The stochastic process approach
5. The Fredholm approach
6. Numerical integration
7. Estimation problems in nonstationary autogressive models
8. Estimation problems in noninvertible moving average models
9. Unit root tests in autoregressive models
10. Unit root tests in moving average models
11. Statistical analysis of cointegration
12. Solutions of problems

Readership: Postgraduate students, teachers and researchers in time series, econometrics and the mathematics of finance

This book is concerned in the main with limiting distributions for statistics which arise in testing for non-standard properties of statistical models: the motivating examples are test statistics for parameter constancy in a state space model and for a unit root in a moving average model. The methods of stochastic calculus are central to this. The chapters introducing Brownian motion, Ito calculus, and functional central limit theory are, without compromising correctness, the most accessible I have yet encountered. Anyone at present nervously using stochastic calculus will be able to apply Girsanov's theorem with a new confidence after working through these chapters. The text is quite dense and requires close study. Details are often left as problems, but every problem is given a full solution. These solutions occupy one hundred and forty pages. I have ordered a copy for the library and recommend that you do the same.

Contents:
Introduction
1. Aims and statistical tools
2. Pareto-type models
3. The extreme value index
4. Weibull type distributions
5. Actuarial applications

The flavour of the book can be summarized as fairly traditional, in that some fairly old-established methods of estimation are proposed, and their asymptotic properties are investigated using the powerful technical apparatus of extreme value theory. Some clue to the style can be gained from the authors' assertion in the Preface that "Our own interest in this subject has evolved gradually from theory to methods". The main methods of estimation are based on plots, such as those quantiles and residual mean life, and the Hill estimator. There seems to be only one brief mention of a likelihood method, though I could have missed others, there being no index. Modern statistical methodology, such as Markov chain Monte Carlo methods, are given little coverage which seems to be confined to univariate independent and identically distributed observations, thus omitting much recent work in the field.

Contents:
1. Introduction
2. Discrete-time Markov chains
3. Monotone Markov chains
4. Continuous-time Markov chains
5. Birth-death processes
APPENDIX A: Review of Matrix Theory
APPENDIX B: Generating Functions and Laplace Transforms
APPENDIX C: Total Positivity

Its title notwithstanding, the book concerns Markov chains, with the focus being on transient behaviour. The treatment is rather deeper than is usual in introductory texts on stochastic processes, although readers are often spared the full glory (or gory) of complete generality. There is a nice blend of classical and modern flavours, the latter including rates of convergence to stationarity, quasi-stationarity, algorithms and numerical methods, and structural results. The treatment is almost entirely algebraic. While this has pedagogical attractions, one disappointment in an otherwise valuable book is the absence of other approaches, perhaps most notably coupling and sample path treatments.

Contents:
1. Introduction
2. Overview of astronomy
3. The character of astronomical data
4. Overview of statistics
5. Resampling methods
6. Spatial statistics
7. Linear regression
8. Multivariate classification and analysis
9. Time series analysis
10. Censoring and truncation
11. Some astronomical controversies

This is not an introductory book on statistics for astronomers. Its aim is to act as a bridge between statisticians and astronomers, giving statisticians an insight into astronomical data and the astronomers an idea of statistical techniques that could be useful to them. The statistical content ranges from elementary exploratory data analysis to sophisticated techniques which are active areas of statistical research such as the analysis of irregularly-spaced time series. Extensive references both to the astronomical and statistical literature are given for each topic as well as information on what software is available.
Inevitably in a book which attempts to de-scribe such a wide range of techniques there are some inaccuracies of description, and grammatical errors abound. However, the authors have made a valuable contribution by alerting astronomers to what statisticians have to offer them and by alerting statisticians to the challenging research problems astronomical data present.

Contents:
1. Procedures and their properties
2. The Bayes approach
3. The empirical Bayes approach
4. Performance of Bayes procedures
5. Bayesian computation
6. Model criticism and selection
7. Special methods and models
8. Case studies
APPENDIX A: Distributional Catalogue
APPENDIX B: Software Guide
APPENDIX C: Answers to Selected Exercises

This book gives an excellent exposition of Bayes and empirical Bayes methods, and links these methods to several applications in the health sciences and other fields. The book gives a well-balanced mathematical and computational treatment of Bayes and empirical Bayes paradigms, and nicely examines the similarities and contrasts in the two approaches. Throughout the book, various modelling techniques arising from real experiments are presented. Models include linear models, generalized linear models, random-effects models, non-linear models, survival models and non-parametric models. Informative and non-informative prior elicitation is nicely presented. Data analyses using Bayes and empirical Bayes methods are presented throughout. Computational strategies involving the EM algorithm, Laplace methods, Markov chain Monte Carlo methods and other related iterative algorithms are discussed and demonstrated for the various models throughout the text.

Contents:
1. The Bayesian approach to statistical archaeology
2. Outline of the approach
3. Modelling in archaeology
4. Quantifying uncertainty: The probability concept
5. Statistical modelling
6. Bivariate and multivariate distributions
7. Bayesian inference
8. Implementation issues
9. Interpretation of radiocarbon results
10. Spatial analysis
11. Sourcing and provenancing
12. Application to other dating methods
13. The way forward

Readership: Students of archaeology, teachers of statistics courses in search of examples and illustrations

Archaeologists face a rich array of difficult inferential problems featuring a paucity of data, competing or conflicting theories and paradigms, and a need to synthesize different kinds of evidence and ex-pert knowledge. In dating artifacts, for example, there may be a need to combine evidence from radio-carbon analysis, tree-ring counts, metal-isotope analysis, and competing theories about the societies that produced the artifacts. This book makes the case that Bayesian methods offer an ideal framework for making inference in archaeological problems.
The leisurely and repetitive introduction to Bayesian statistical analysis will be most appropriate for students with little statistical or mathematical background; the wealth of case studies and examples will make the book a useful occasional source for those teaching elementary applied statistics courses to broader social-science audiences. The statistical methods presented are contemporary, for example, MCMC implementation is discussed for non-conjugate priors, but non-technical, for example, reference priors other than the uniform are only alluded to and never used.

Contents:
PART I : A Survey of Models
1. A survey of models for discrete-valued time series
PART II: Hidden Markov Values
2. The basic models
3. Extensions and modifications
4. Applications
APPENDIX A: Proofs of Results Used in the Derivation of the Baum-Welsch Algorithm
APPENDIX B: Data

Discrete-valued time series are common in practice, but the methods for their analysis are not well known. This is the first book which summarizes the recent developments in this area. The authors have aimed at applied statisticians as their primary reader-ship. The first chapter provides a brief survey of existing models, other than the hidden Markov model, as data-analytic tools for analyzing discrete-valued time series. These include, among others, discrete ARMA models, models based on thinning, bivariate geometric models, Markov regression models, parameter-driven models and state-space models. Although the discussion on various models is of more probabilistic interests than statistical, references on the applications are included. The remaining three chapters are devoted to one topic, a class of hidden Markov time series models This class seems versatile enough to cater for the various applications. Since each model has a simple structure, computationally tractable algorithms have been developed to evaluate the estimates of parameters. The book has included applications to nine real sets of data drawn from a wide range of disciplines. The publication of this book is a welcome and timely reminder that we still know too little about the subject and it is hoped that it will provoke further research in this important area.

Contents:
Introduction
PART I : Pure Inverse Problems
PART II : Linear Inverse Problems with Noise
PART III : General Linear Models
PART IV : A System of Economic Statistical Relations
PART V : Linear and Nonlinear Dynamic Systems
PART VI : Discrete Choice-Censored Problems
PART VII : Computational Notes
PART VIII: Epilogue
Readership: Econometricians, statisticians, quantitative economists

In economics and more generally in the social sciences many models contain unknowns that are not accessible to direct measurement. Faced with these problems the book develops a non-linear inversion procedure that provides a basis for information recovery about unknown and unobservable numbers, vectors or functions. While the approaches taken in the book are based on the information-theoretic concept of entropy minimization, this is a book about econometrics and not about information theory. The authors consider both static and dynamic models and treat a variety of explicit examples. The focus of the book is clearly on applications. It ends with a chapter on computational issues where even explicit computer code is provided.

Contents:
PART I : Introduction
1. Introduction
2. Simple illustrative and motivating examples
3. Empirical distributions: Statistical laws in macroeconomics
PART II : Modeling Interactions
4. Modeling interactions I: Jump Markov processes
5. Modeling interactions II: Master questions and field effects
6. Modeling interactions III: Pairwise and multiple-pair interactions
PART III: Hierarchical Dynamics and Critical Phenomena
7. Sluggish dynamics and hierarchical state spaces
8. Self-organizing and other critical phenomena in economic models
Elaborations and future directions of research

Readership: Specialists in stochastic processes and their applications in economic theory

This book describes a useful toolkit of analytical techniques in stochastic processes. Most of the economic applications are in Chapters 4 and 5, where economic agents are modelled as jump Markov processes. The book's main contribution is to show how properties of aggregates of such agents can be derived analytically from their individual laws of motion. Field effects refer to situations in which the transition probability density function for an agent depends on some property of the aggregate. For example, the results of a firm's decision to invest in new equipment may depend on how many other firms are doing so. The Chapman-Kolmogorov equations are then solved ingeniously to show implied dynamics for the aggregate collection of agents. It may be too soon to describe this as macro-economics, for there is no general equilibrium of the economy here. However, this book contains a wealth of suggestive mathematical examples which may have applications in labour economics or industrial organization.

Table des matières
1. Modèles Markoviens
2. Des algorithmes stochastiques: Pourquoi? comment?
3. Convergence d'algorithmes à pas décroissants
4. Cibles attractives
5. Vitesses en grandes déviations
6. Recuit simulé sur un espace fini
7. Recuit simulé vectoriel

Cet ouvrage présente un traitement de la théorie et applications des algorithmes stochastiques. La forme originale du livre est un cours pour étudiants de niveau troisième cycle universitaire. Ceci fait que les éléments nécessaires de la théorie des probabilités et les processus stochastiques sont repris et qu'il y a des exercises nombreuses dans le texte. Le choix des sujets est très riche et les applications sont nom-breuses: automatique, images, neurones, ... . Chaque chapitre se termine avec une liste précieuse de références à la littérature.

Contents:
Introduction
1. Basic definitions
2. Inequalities
3. Law of large numbers
4. Weak convergence
5. Functional limit theorems
6. Approximation estimates
7. Asymptotic expansions
8. Large deviations
9. Law of iterated logarithm
10. Dependent variables

Since the fundamental paper of Hoeffding in 1948, the class of U-statistics, generalizing the usual sample mean, has been studied in many papers and more recently also in books. This is yet another book on this topic. It contains a lot of material that was available in other books, but this time there is special focus on the case where the kernel function of the U-statistic is Banach-valued. The author calls them UB-statistics. The structural (martingale) and asymptotic properties (law of large numbers, weak convergence, expansions, large deviations, ...) are discussed in the form "theorem plus proof". Also the case of dependent variables is dealt with in the last chapter. The bibliography has about four hundred and fifty items and of about one hundred of them, Borovskikh is the author or co-author. There are also (incomplete) bib-liographical comments at the end of the book. A few criticisms on the monograph are the following. First is the fact that many pages are devoted to the classical (real-valued) U-statistics results. Secondly, the probabilistic theory for B-valued U-statistics is beautiful, but many readers will ask for some motivation for treating the material in this generality. Thirdly, there is no author index and the subject index is only two pages long. Researchers in probability theory and mathematical statistics will of course like to have this book in their library.

Contents:
1. Introduction
2. Discrete processes
3. Continuous processes
4. Pricing market securities
5. Interest rates
6. Bigger models

The book starts with a description of pricing derivative securities in a discrete economy. A brief description is given of continuous time processes, Itô calculus, the Martingale representation theorem and how to replicate a derivative security. The last three chapters are devoted to describing how to price and hedge different types of derivatives in a continuous time context. The authors state that the reader is not expected to have any particular prior body of knowledge except for some (classical) differential calculus and experience with symbolic notation. A reader without knowledge of probability theory will find this book challenging. It is perhaps best suited for an elective course in applied mathematics for undergraduates.

Contents:
1. Financial derivatives: A brief introduction
2. A primer on the arbitrage theorem
3. Calculus in deterministic and stochastic environment
4. Pricing derivatives
5. Tools in probability
6. Martingales and martingale representations
7. Differentiation in stochastic environments
8. The Wiener process and rare events in financial markets
9. Integration in stochastic environments: The Ito integral
10. Ito's lemma
11. The dynamics of derivative prices
12. Pricing derivative products. Partial differential equations
13. The Black-Schols PDE. An application
14. Pricing derivatives products. Equivalent Martingale measures
15. Equivalent Martingale measures application
16. Tools for complicated derivative structures

The material is presented in a very intuitive and heuristic way. The reader will get a good feeling why the theory works. However, mathematics are presented with a lot of hand-waving. Let me give two examples. The conditions on the existence of the Ito integral are, although sufficient, misleading. The conditions are not in general terms and only applicable to the case of stochastic differential equations. But in this case the reader needs a lot of know-how in order to apply the L2-condition. Stochastic differential equations are treated via the Euler approximation scheme. For the Black-Scholes model, this means that negative values can be encountered. This fact is not even mentioned. Also the convergence problem is not mentioned. When the square-root process is treated, a numerical illustration is given and the author says that the reader can see it looks the same as for the Black-Scholes example. The use of the square-root process as an alternative asset price model is then mentioned but the fact that this process attains the value zero with certainty is over-looked.
The theory is always illustrated with the Black-Scholes model. Since other processes are also mentioned, the reader might get the impression that it always works as easily. The example of the square-root process shows that surprises can occur.
The book is good as an introduction to the field. The author could have given somewhat more words of warning when over-simplifying mathematics.
The book is suited for practitioners, who are not that much interested in exact mathematics. Acade-mics working in the field of mathematical finance should read the book with care and criticism.

Contents:
1. Multivariate probability
2. Construction sequences
3. Propagation in join trees
4. Resources and references

Readership: Scholars and students of artificial intelligence, operations research, and applied statisticians who use probabilistic methods

This monograph consists of the first three chapters of what was to have been a longer work covering the material the author had presented in a series of lectures. Unfortunately (?), the author was distracted into writing another book, 'The Art of Causal Conjecture' (MIT Press, 1996) and never completed the present one. However, since the discussion of join-tree architectures presented in these three chapters is still missing from the literature, he decided to publish these chapters alone. They are supplemented by exercises and by a chapter on 're-sources', which also gives an annotated bibliography and details of software.
The monograph describes how modularity in a probabilistic model enables computation of probabilistic inferences, and outlines the main architectures for capitalizing on this modularity. It provides a good, brief, but rigorous introduction to the ideas.

Contents:
1. Mathematical preliminaries
2. Optimization in Rn
3. Existence of solutions: The Weierstrasse theorem
4. Unconstrained optima
5. Equality constraints and the theorem of Lagrange
6. Inequality constraints and the theorem of Kuhn and Tucker
7. Convex structures in optimization theory
8. Quasi-convexity and optimization
9. Parametric continuity: The maximum theory
10. Super modularity and parametric monotonicity
11. Finite-horizon dynamic programming
12. Stationary discounted dynamic programming

The author of this text has paid particular attention to his aim of presenting a clear but cohesive blend of theory with applications. The presentation has five characteristics. First, every result is accompanied by a complete proof, which is usually located at the end of the chapter. Secondly, immediately after the statement of a result, examples are given that illustrate why the result would be invalid were any of the assumptions not satisfied. Thirdly, there is a section which discusses procedures using the result to determine an optimum. Fourthly, the chapter's results in applications are illustrated by the solution of numerical examples. The examples come from economic analysis, for example maximizing a utility function subject to budget conditions. Finally, at the end of each chapter, there are exercises which cover both the theory and applications. No solutions are given. This text is to be praised for its clarity. The author's style and presentation inspires confidence so that as a reader you always know where you are. To either teacher or student, I highly recommend this text.

Contents:
1. Introduction
2. Causal and Bayesian networks
3. Building models
4. Propagation in Bayesian networks
5. Use of Bayesian network models
6. Actions
APPENDIX A: Construction of Junction Trees (Proofs)
APPENDIX B: Value of Information (Proofs)

Readership: Artificial intelligence students, or statistics teachers and students

This humourously-written book is "intended for both classroom use and self-study, and it addresses persons who are interested in exploiting the Bayesian network approach for the construction of decision-sup-port systems or expert systems." The approach is that of modelling reasoning under uncertainty and will appeal to students of artificial intelligence; for statistics students, these networks are causal probabilistic networks being "Bayesian" by use of Bayes' Theorem. Only random variables having a finite number of possible values are considered.
Examples are used throughout to motivate and illustrate the modelling approach. The mathematical background is elementary calculus, graph theory and probabilistic theory. Exercises and a restricted version of the impressive commercial program HUGIN en-courage the reader to construct their own models. How-ever, with no documentation in the text, and only cursory help on the disk, readers unfamiliar with HUGIN will require some time exploring the software before it becomes as useful as intended; the tutorial introduction is "under construction".
The book provides a useful introduction to a relatively new area and could form the basis of an interesting upper year undergraduate course.

Table des Matières:
PARTIE I : L'Analyse des Données
Introduction
1. La méthode ACP (Analyse en Composantes Principales)
2. La méthode AFC (Analyse Factorielle des Correspondances)
PARTIE II : Les Méthodes Exploratoires d'Analyse des Donnés Evolutives
3. La méthode STATIS (Structuration de Tableaux a Trois Indices de la Statistique)
4. La méthode AFM (Analyse Factorielle Multiple)
5. La méthode DACP (Double Analyse en Composantes Principales.
6. Comparison des méthodes. Selection méthodes d'une méthodologie
PARTIE III: Les Méthodes de Classification en Analyse de Données Evolutives
Introduction
7. La méthode de classification des centres mobiles
8. Quelques distances en analyse des donnés évolutives

Readership: Students and research workers interested in 3-mode multivariate methods

The first two chapters recapitulate the theories of principal components and correspondence analysis. The novelty is in the remaining chapters concerned with the theory and application of méthodes évolutives concerned with the simultaneous (conjoint) analysis of sets of multivariate data-matrices, perhaps collected chronologically but not necessarily so. The same variables and objects may be observed for each set (DACP), or the objects remain the same but the variables differ (STATIS and AFM) or the variables differ but the objects remain the same (dual of STATIS). A variant of DACP, ASCM (l'Analyse de Séries Chrono-logiques Multidimensionelles) seems to be the only method discussed which specifically considers sequential information. References are mainly to French work with little or no discussion of the extensive related "Anglo-Saxon" literature. The book will be useful to those who apply these methods or to statisticians who wish for an introduction to French developments in the analysis of sets of data matrices.

Contents:
1. Introduction and overview
2. Likelihood
3. Maximum likelihood estimation
4. Hypothesis testing
5. Linear models
6. Generalised linear models

From the preface: "The aim is to show how the main body of currently used statistical techniques can be generated from a few key concepts, in particular the likelihood." The author achieves this aim extremely well. Purely mathematical niceties are not glossed over, but are not given undue importance. The only improvement I can suggest for a likely second edition is to introduce maximum likelihood estimation via Godambe's results, that älogL/äÈ=0 is, in the regular case, in finite samples, the essentially unique optimum unbiased estimating equation. The treatment using asymptotics tends to suggest to students that maximum likelikhood estimation is superior to other methods only in large samples.

Contents:
Prologue: A Living Legend
1. The formative years - Early childhood and schooling
2. Putting down some roots - Andhra University
3. In search of a job
4. At home in the Indian Statistical Institute
5. Mice and skeletons - at Cambridge
6. Back to Calcutta - Professor at 28
7. Prof. Dr. Dr. h.c. mult.
8. A plethora of interests
9. Anecdotal insights: by and about C.R. Rao
Epilogue: Conversations with C.R. Rao
APPENDIX 1: C.R. Rao's Epigrams on Statistics
APPENDIX 2: Contributions to Statistics
APPENDIX 3: Biographical Details

Readership: Statisticians, mathematicians, probabilists, scientists, students, teachers

"By the age of 28, Radhakrishna [C.R. Rao] was the author of forty published papers, with several results to his credit which would eventually be named after him: the Cramér-Rao inequality, Rao-Blackwell theorem, Lehman-Scheffé-Rao theorem, Rao's Score test, Neyman-Rao test, Rao's F-test, Hamming-Rao bound, .... After his return to Calcutta, he continued producing: Fisher-Rao theorem, Rao's canonical factor analysis, Rao's quadratic entropy, Minque theory, Rao's g-inverse of a matrix, Rao-Rubin, Kagan-Linnik-Rao and Lau-Rao-Shanbhag theorems."
This very readable biography chronicles the life and many achievements of this "one hundred percent Indian bred" and world-renowned statistician. Its strength lies in showing the human dimension of this unique researcher, teacher and inspirer. It places him in his cultural context. It chronicles his family back-ground and childhood, his student days at Andhra, Calcutta and Cambridge Universities, and the great and diversified influence that he has had on the Indian Statistical Institute, at the University of Pittsburgh, at Pennsylvania State University and world-wide.
It also gives brief descriptions of his statistical researches in an appendix; much fuller details are in his autobiographical account of his research work, Statistics as a Last Resort, issued by Statistics '91 Canada, Concordia University, Montreal. Lists of his academic qualifications, very many honorary degrees, fellowships, medals, cash awards, fest-schrifts, special journal issues, editions of his books, and nearly fifty successful Ph.D. students are provided. The anecdotes by Rao himself and by his colleagues and former students show him as a "brilliant, disciplined, sincere, meticulous, focused, energetic, enthusiastic, humorous, approachable and unassuming" person.
The lengthy conversation with his biographer complements and supplements his well-known conversation with M.H. DeGroot in Statistical Science, (1987) Vol. 2, pp. 53-67.

Contents:
1. Preliminaries
2. Clinical trials and research
3. Ethical considerations
4. Clinical trials as experimental designs
5. Bias and random error
6. Objectives and endpoints
7. Sample size and power
8. The study cohort
9. Treatment allocation
10. Data-dependent stopping
11. Counting patients and events
12. Estimating clinical effects
13. Prognostic factor analysis
14. Reporting
15. Factorial designs
16. Cross-over designs
17. Overviews/meta-analyses
18. Fraud and misconduct
19. Data and programs
20. Notation and terminology

"The best time to contemplate the quality of evidence from a clinical trial is before it begins." From this fine opening statement, this book just gets better and better. The book assumes a working knowledge of basic biostatistics and some familiarity with clinical trials. However, it is so well written, and referenced, that no one should be put off by this. If you have any interest in clinical trials, get this book.
The comprehensiveness of the book is well illustrated by the chapter titles but there are many gems in the subheadings as well. I was particularly pleased to see "5.2.5 Using a one- or two-sided hypothesis test is not the right question" and "5.2.8 Post Hoc Power calculations are not helpful". Not everything is covered. The oncology background of the author is sometimes evident. Bayesian methods receive minimal, although sympathetic, treatment. However, the author's aim throughout is clearly illustrated by the comment "Although differences between frequentist and Bayesian methods are consequential, there are more important methodologic concerns in clinical trials."
It will be a shame, and I will be surprised, if this does not very quickly become a standard reference for clinical trial methodology.

Contents:
1. The First Principle
1.2 The law of likelihood
1.3 Three questions
1.4 Towards verification
1.5 Relativity of evidence
1.6 Strength of evidence
1.7 Counterexamples
1.8 Testing simple hypotheses
1.9 Composite hypotheses
1.10 Another counterexample
1.11 Irrelevance of the sample space
1.12 The likelihood principle
1.13 Evidence and uncertainty
2. Neyman-Pearson Theory
2.2 Neyman-Pearson statistical theory
2.3 Evidential interpretation of the results of Neyman-Pearson decision procedures
2.4 Neyman-Pearson hypothesis testing in planning experiments: Choosing the sample size
3. Fisherian Theory
3.2 A method for measuring statistical evidence: The test of significance
3.3 The rationale for significance tests
3.4 Troubles with p-values
3.5 Rejection trials
3.6 A sample of interpretations
3.7 The illogic of rejection trials
3.8 Confidence sets from rejection trials
3.9 Alternative hypotheses in science
4. Paradigms for Statistics
4.2 Three paradigms
4.3 An alternative paradigm
4.4 Problems of weak and misleading evidence: Normal distributed mean
4.5 Understanding the likelihood paradigm
4.6 Evidence about probability
5. Resolving the Paradoxes from the Old Paradigms
5.2 Why is a power of only 0.80 OK?
5.3 Peeking at data: Repeated tests
5.4 Testing more than one hypothesis
5.5 What is wrong with one-sided tests?
5.6 Why not use the most powerful test?
5.7 Must the significance level be predetermined? And is the strength of evidence limited by the researcher's expectations?
6. Looking at Likelihoods
6.2 Evidence about hazard rates in two factories
6.3 Evidence about an odds ratio
6.4 A standardized mortality ratio
6.5 Evidence about a finite population total
6.6 Determinants of plans to attend college
6.7 Evidence about probabilities in a 2x2x2x2 table
6.8 Evidence from a community intervention study of hypertension
6.9 Effects of sugars on growth of pea sections: Analysis of variance
7. Nuisance Parameters
7.2 Orthogonal parameters
7.3 Marginal likelihoods
7.4 Conditional likelihoods
7.5 Estimated likelihoods
7.6 Profile likelihoods
7.7 Synthetic conditional likelihoods
8. Bayesian Statistical Inference
8.2 Bayesian statistical models
8.3 Subjectivity in Bayesian models
8.4 The trouble with Bayesian statistics
8.5 Are likelihood methods Bayesian?
8.6 Objective Bayesian inference
8.7 Bayesian integrated likelihoods
APPENDIX: The Paradox of Ravens
Each chapter has an introduction and summary. All but Chapter 6 are followed by exercises.

Well-chosen examples of real data show the logical weakness of many traditional accounts of statistical inference and the value of looking at the likelihood function whenever we can. It is shown that by doing this we gain many advantages claimed for the Bayesian approach, without incurring the disadvantage of its subjectivity. Careful reading is strongly recommended.

Contents:
1. Measurement: Problems and strategy
2. Validity and reliability
3. Labor wastage
4. Inequality
5. Mobility
6. Price levels
7. Population level and individual-level measures
8. Common scaling of individuals and items
9. Individual-level management: General theory
10. Principles and problems of implementation
11. Extensions and limitations

Measurement is a curiously neglected study. Despite the dependence of scientific work on quantitative methods, the basis of quantification is seldom made explicit. This book is a welcome contribution to the field. The first chapters are particularly import-ant for the new researcher in stressing that measurement is problematic, that it always has a purpose and requires a strategy. If one is to rely on measures they must be both valid and reliable. The remainder of the book guides the reader steadily through population and individual levels of measurement. Each aspect of the developing argument is illustrated by a sustained and important example. Written for social scientists, this book could equally usefully be read by managers, and in practice their statistical advisers, who all too often rely on spurious "measures" in attempting to control only too human organizations.

Contents:
Introduction
1. Setting the stage
2. The usual suspects
3. How habits of mind govern intuition
4. The risk matrix
5. Experts and victims
6. Examining cases
7. Two modest proposals: Some background
8. "Do no harm"
9. Political externalities
10. Afterword

Readership: Risk assessors and managers, risk communicators, intellectual leaders of citizens' groups opposing experts' views on technological/environmental risks

The subtitle of this book, "Why the public and experts disagree on environmental issues", is more in-formative than the title. Margolis discusses the various origins for disagreements between experts and environmentalist advocates on the risk of technological interventions that predominate in the literature on the discord. These include: (1) differences in opinions as to what and whose ends ought to be served, (2) public distrust in the agencies that claim that the risk is under control, and (3) different meanings (denotative and connotative) attached to such terms as "risk" and "riskiness". It is argued by Margolis, however, that these three sources of disagreement do not suffice to explain the persistent controversies. He proposes that people as experts and people as lay persons use different habits of mind arriving at the conclusions they believe in, habits as deeply engrained as tying granny or straight knots. Readers who remain unconvinced de-spite the lengthy arguments presented in this somewhat heavy-plodding text may well run the risk of being viewed as evidence for the author's proposition.

Contents:
PART I : The New Paradigm
PART II : Fractal Structure in the Capital Markets
PART III: Nonlinear Dynamics
PART IV : Living with Complexity

The first edition of this book was instrumental in making the topic of chaos popular among financial analysts, the main implication being that one ought to go beyond linear dynamics in order to capture the more subtle behaviour of financial data. The second edition contains essentially the same material, various small editorial matters have been taken care of. The main addition is to be found in Part IV on Complexity Theory. Also, the added diskette (VBASIC) has been updated. It contains four sets of data, two financial ones, the sunspot data and the river Nile data, together with some introductory programmes. The style of writing is very informal. As the author states, the book is intended to communicate the concepts behind fractals and chaos theory as they apply to the capital markets and economics. The professional statistician or indeed mathematical chaos specialist will not find a lot of exciting material in it, but then they do not really belong to the intended reader-ship.

Contents:
1. Introduction
2. Linear and nonlinear processes
3. Univariate ARCH models
4. Estimation and tests
5. Some applications of univariate ARCH models
6. Multivariate ARCH models
7. Efficient portfolios and hedging portfolios
8. Factor models, diversification and efficiency
9. Equilibrium models

Readership: Statistics and econometrics specialists, particularly those who are familiar with theoretical and empirical work in finance

This expanded version of a French language original surveys ARCH models from the viewpoint of theory in statistics and finance, along with brief discussions of some empirical findings. It is written in textbook form with a minimum of references to specific sources in the body of the text so that the narrative flows smoothly (but tersely). Extensive lists of references, grouped by topic, are given for each chapter; these would be better described as further readings in most cases. All chapters except Chapters 1, 6 and 9 conclude with exercises. Chapter 4 emphasizes pseudo maximum likelihood estimation but it also includes a shorter discussion of a two-step procedure. While the grouping of references is most helpful, more careful editing would have caught some missing and some misplaced references.

Contents:
34. General remarks
35. Multinomial distributions
36. Negative multinomial and other multinomial-related distributions
37. Multivariate Poisson distributions
38. Multivariate power series distributions
39. Multivariate hypergeometric and related distributions
40. Multivariate Pólya-Eggenberger distributions
41. Multivariate Ewens distribution
42. Multivariate distributions of order s
43. Miscellaneous distributions

A single chapter of the 1969 edition has through expansion and updating matured into an entire volume. Only the new edition on continuous multivariate distributions is still to come in this series. The original policy of providing facts about commonly used distributions is stretched here. The index lists three columns of them. The authors acknowledge a review by Papageorgiou in Encyclopedia of Statistical ScienceC Update Volume 1 (due this year) and the book Bivariate Discrete Distributions by S. & K. Kocherlakota (1992) [Short Book Reviews, Vol. 12, p.42]. The glossary, list of notations, and the mechanism for naming the distributions would in themselves make this an essential source for all authors in this area. Any facile comment that this book would settle pub arguments about the difference between the generalized binomial multivariate Poisson distribution and the Poisson-negative hypergeometric is misplaced. It is too impressive for that. The authors do not despair at these permutations and keep the writing sensitive to the readership. They recognize that since 1969 emphasis has moved towards more realistic modelling rather than to the development of estimation methods for the classical models, except in the case of Bayesian statistics.
There are forty-six areas of application given in the index. Of particular note is the multivariate Ewens distribution, of importance both in genetics and in Bayesian methodology.
All praise is due to the typists and technical editors.

Sommaire:
PARTIE 1: Traitement d'un fichier de notes
PARTIE 2: Eléments remarquables et éléments aberrants
PARTIE 3: Fiches techniques

Voici un cours de statistiques très original: partant de données réelles, les auteurs introduisent progressivement les techniques de la statistique. Le fichier de données utilisé dans ce livre comporte les résultats obtenus au bac et pendant l'année scolaire pour 993 élèves de classe terminale scientifique. L'étude commence avec le "nettoyage" des données et avec une description des actions possibles contre les erreurs. Aussi le délicat problème de données manquantes est discuté. Après on commence avec la statistique descriptive pour une ou plusieurs variables discrètes ou continues. Une présentation plus systématique des aspects techniques de l'analyse statistique et dans la dernière partie du livre. Cet ouvrage est certaine-ment intéressant pour un public très large parce qu'il est accessible sans mathématiques.

Contents:
1. Introduction
2. Elements of probability
3. Statistical inference
4. Off-line quality control
5. Shewhart control charts
6. More general control charts
7. Sampling inspection by attributes
8. Sampling inspection by variables: A loss function approach

Readership: Students seeking a first exposure to statistical methods useful in quality control
This is an introductory textbook which aims to integrate the traditional basics of probability and statistics with off-line and on-line quality control. The book covers only a small selection of topics from these vast fields, and each of these topics is treated only very lightly. Although the brevity of the text may help the reader to appreciate the "essence" of each subject, students will not be well prepared for industrial application of any of these methods. The chapter on experimental design is particularly weak. All of the important ideas in the book can be studied in greater depth elsewhere.

Contents:
1. Basic probability and statistics
2. Linear regression analysis
3. Variance components and process sampling design
4. Measurement capability
5. Introduction to statistical process control
6. Statistical process control implementation

This book provides sound advice on effecive application of many commonly used, and some less commonly used, statistical methods in industrial practice. Its principal strengths are the author's obviously extensive experience as a practising statistician in the semiconductor industry and a number of stimulating and instructive examples from that area. The chapter on measurement capability is a particularly useful remind-er of the importance of reliable data. SASR routines for some of the methods discussed are included in an appendix. In view of the large number of topics addressed in this book, it is not surprising that some receive inadequate attention. The organization of the book into only six chapters leads to some curious combinations of topics within a chapter. However, these relatively minor distractions diminish only slightly the potential utility of this book for technical practitioners in the manufacturing industry, especially those in the semiconductor sector.

Contents:
1. Examples of survival data
2. Basic quantities and models
3. Censoring and truncation
4. Nonparametric estimation of basic quantities for right censored and left truncated data
5. Estimation of basic quantities for other sampling schemes
6. Topics in univariate estimation
7. Hypothesis testing
8. Semiparametric proportional hazards regression with fixed covariates
9. Refinements of the semiparametric proportional hazards model
10. Additive hazards regression models
11. Regression diagnostics
12. Inference for parametric regression models
13. Multivariate survival analysis
APPENDIX A: Numerical Techniques for Maximization
APPENDIX B: Large Sample Tests Based on Likelihood Theory
APPENDIX C: Statistical Tables
APPENDIX D: Selected Data Sets

Over the past few years a number of books on survival analysis have appeared, some directed at professional or research statisticians and others at less sophisticated practitioners. This book differs from all of these in that it brings survival analysis right into the class room, offering an excellent course in survival analysis for Masters-level students or indeed for statisticians who wish to extend their knowledge of this subject.
The book is very well organized. In the first chapter, nineteen examples of survival data are discussed. These are used to illustrate the theory and for the copious exercises that accompany each of the subsequent chapters. The authors treat the subject from a classical point of view and the mathematical level is compatible with that. A brief review of the alter-native development of the subject through counting processes is given in Chapter 3 and further references and discussion are given in the theoretical notes that are part of each chapter.
The subject is developed mathematically, but strong emphasis is placed on the practical implementation of the techniques. Included in each chapter are practical notes that extend the theoretical developments in the text and discuss relevant computer programs.
Students completing a course based on this book would be equipped to handle many types of survival data. Further the book is an excellent source of references to the ever-growing literature on this important
topic.

Contents:
1. Introduction
2. Formulating tests for the presence of immunes and sufficient follow-up
3. Properties of the Kaplan-Meier estimator
4. Nonparametric estimation and testing
5. Parametric models for single samples
6. Use of concomitant information
7. Large-sample properties of parametric models: Single samples
8. Large-sample properties of parametric models with covariates
9. Further topics

Readership: Statisticians in medicine and biometry, epidemiologists and sociologists

This book adds a new perspective to survival analysis. Are there, perhaps, among the population studied, certain members who are immune to the event in question? The possibility that there are such individuals is indicated by censored observations at large survival times.
The authors develop tests for their presence, first from a non-parametric view based on the Kaplan-Meier estimator and then from a parametric view based on exponential and Weibull mixture distributions. The parametric tests are extended to cover grouped data and covariates. In the final chapter, they discuss how their methods compare and contrast with other aspects of survival analysis such as methods-based modelling of the hazard function and competing risks. Lastly they show how to estimate the censoring distribution and the probability of being immune.
While acknowledging that martingale theory is at the basis of their methods, the mathematical level is maintained at one that would be accessible to most statisticians.
A most attractive feature of the book is the carefully worked examples using data drawn not only from traditional medical applications, but also from criminology. The examples are detailed enough for the most unsophisticated user to follow, and the results of the procedures are interpreted in context of the study.
All in all, the authors have presented a use-ful additional technique to survival analysis, and their lucid exposition should ensure that it becomes part of the standard methodology. This book is a welcome addition to the growing literature on survival analysis.

Contents:
PART I : Fundamentals of Pattern Recognition
0. Basic concepts of pattern recognition
1. Decision-theoretic algorithms
2. Structural pattern recognition
PART II : Introductory Neural Networks
3. Artificial neural network structures
4. Supervised training via error backpropagation: derivations
PART III: Advanced Fundamentals of Neural Networks
5. Acceleration and stabilization of supervised gradient training of MLPs
6. Supervised training via strategic search
7. Advances in network algorithms for classification and recognition
8. Recurrent neural networks
PART IV : Neural, Feature, and Data Engineering
9. Neural engineering and testing of FANNs
10. Feature and data engineering
PART V : Testing and Applications
11. Some comparative studies of feed-forward artificial neural networks
12. Pattern recognition applications
Readership: Course text for senior undergraduate or graduate students in computer science, electrical engineering, or computer engineering studying pattern recognition or neural networks, reference and resource for researchers and professionals

This is a fairly comprehensive introduction to feed-forward neural networks, including coverage of multilayer perceptrons, functional link nets, and radial basis functions. The book discusses both supervised and unsupervised classification, and even has a (some-what old-fashioned) chapter on structural methods. The algorithms are presented in pseudocode and there are many illustrative elementary numerical examples. The book has exercises at the end of each chapter, and also puts the references there.
The emphasis and style is on algorithms rather than principles, perhaps befitting the intended computer science/engineering audience. However, this means that it has a rather superficial discussion of theoretical issues, such as, for example, the topics of overfitting and generalizability, which are now recognized to be of fundamental importance. Because of this, statisticians may well find more appeal in the books by Bishop Neural Networks for Pattern Recognition, Oxford, 1995; Ripley, B.D. Pattern Recognition and Neural Networks, CUP, 1996 [Short Book Reviews, Vol 16, p.29], or (if I may) Hand, D.J. Construction and Assessment of Classification Rules, Wiley, 1997, [Short Book Reviews, Vol. 17, p.46]. However, the book is accessible and would be well-suited to serve as a text for its intended audience.

Contents:
PART I : Basic Ideas
1. Introduction
PART II : Constructing Rules
2. Fisher's LDA and other methods on covariance matrices
3. Non-linear methods
4. Recursive partitioning methods
5. Nonparametric smoothing methods
PART III: Evaluating Rules
6. Aspects of evaluation
7. Misclassification rate
8. Evaluating two-class rules
PART IV : Practical Issues
1. Some special problems
2. Some illustrative applications
3. Links and comparisons between methods

Readership: Statisticians and computer scientists with a practical interest in classification

This book provides a much needed guide through the burgeoning field of classification. Its particular value is the emphasis it places on choosing a classifier according to the features of the problem. It also draws attention to the problem of assessing the performance of a classifier in a given situation, again stressing that, in seeking the "best" classifier, there may be more than one interpretation of "best", according to the demands of the application in hand. Al-though clear and concise, and containing much basic material, it is not really a book for the novice. It could and should be read in conjunction with other introductory texts. The practical examples are most welcome.

Contents:
1. Introduction
2. Reduction to dimensionality
3. Development and study of multivariate dependencies
4. Multidimensional classification and clustering
5. Assessment of specific aspects of multivariate statistical models
6. Summarization and exposure

The first edition of this book, published in 1977, was truly ground-breaking with its innovative and refreshingly practical view of multivariate analysis. In the intervening twenty years there has been much new research into practical, computational multivariate methods, and many practically-oriented texts are now available, so production of a second edition must have been quite challenging. The author neatly side-steps some of this challenge, stressing in the preface that he will focus exclusively on "classical" techniques and ignore the "new paradigms", such as smoothing, non-parametric function fitting, data resampling, etc. The main additions thus come in the sections dealing with classification, clustering summarization and robust estimation, and there is an expanded appendix on soft-ware. However, the revisions seem disappointingly limited: only about thirty post-1977 references, and no mention of such important descriptive/graphical techniques as biplots, correspondence analysis, projection pursuit or Procrustes analysis. Much of the non-linear material is also very dated. While the book is undoubtedly still very valuable, it has perhaps lost the sparkle and freshness of its predecessor.

Contents:
Prologue: Why Wavelets?
1. Wavelets: A brief introduction
2. Basic smoothing techniques
3. Elementary statistical applications
4. Wavelet features and examples
5. Wavelet-based diagnostics
6. Some practical issues
7. Other applications
8. Data adaptive wavelet thresholding
9. Generalizations and extensions

Readership: Advanced undergraduates and researchers in mathematics and statistics

The wavelet literature is one of the fastest growing of any topic. Several books are now available in this area, although most are written for a quite sophisticated mathematical audience, with a bias to-wards signal processing. This book concentrates on statistical applications of wavelet methodology, and is written at a fairly basic mathematical level. A large number of different facets of wavelet theory and statistical applications are treated in this slim volume. For example the reader will find wavelets on an interval and boundary handling discussed, along with wavelet thresholding (shrinkage), cross-validation, methods, two-dimensional wavelets and wavelet packets. As a result of the scope of the book, the topics are not treated in depth, but a good basic overview is pro-vided, and this is the strength of the book. It will be most useful to those wishing to find out about the basics of statistical applications of wavelets, rather than those wishing to find answers to specific quest-ions. Overall I would recommend this as a useful source book for a research library.

Contents:
1. Introduction
2. Basic concepts of dependence
3. Fréchet classes
4. Construction of multivariate distributions
5. Parametric families of copulas
6. Multivariate extreme value distributions
7. Multivariate discrete distributions
8. Multivariate models with serial dependence
9. Models from given conditional distributions
10. Statistical inference and computation
11. Data analysis and comparison of models

Readership: Graduate students, lecturers and researchers in statistics, probability and biostatistics

This book provides an in-depth coverage of the development and interpretation of models for multivariate non-normal binary, count and extreme value response data, with covariates, explanatory variates or other factors taken into account. The methods, which are applicable to many important practical problems, particularly in biostatistics and epidemiology, are based on multivariate models using multivariate distributions with univariate margins belonging to a given family. Dependence concepts, more generally useful than correlations, are introduced in Chapter 2. These include positive quadratic dependence and the concordance ordering used with parametric families of copulas, which are multivariate distributions with all univariate margins being uniform on (0,1). One of the important ideas in the book is the use of copulas to summarize the dependence in a multivariate distribution, independent of the univariate margins. Other dependence structures and multivariate situations are covered in considerable depth with, in each case, the modeling process increasing in complexity as each chapter develops. At the end of each chapter there are exercises for the serious reader and a set of unsolved problems for those who wish to research this area. Those whose main desire is to see how the methods and ideas can be applied to practical problems, are recommended to start with Chapter 11 and then select the relevant theoretical sections from earlier parts of the text. This book is not an easy read but contains a wealth of theoretical information about model building in practically relevant multivariate situations.

Contents:
1. Introduction
2. Diagnostics and remedial measures
3. Regression with matrix algebra
4. Introduction to multiple linear regression
5. Plots in multiple regression
6. Transformations in multiple regression
7. Selection of regressors
8. Polynomial and trigonometric terms
9. Logistic regression
10. Nonparametric regression
11. Robust regression
12. Ridge regression
13. Nonlinear regression
14. Experimental design for regression
15. Applications of regression

Readership: Undergraduates and graduates in statistics, mathematics and the engineering and physical sciences

Ryan has set out to write a book telling what is known and not known about various methods of fitting models to data. He picks up a number of current topics, provides recent references, and sifts the claims and counterclaims with great care. "A concerted effort has been made to present state-of-the-art regression methodology." (Preface.) Set within the first seven chapters is a basic course on regression, obtained by picking out a suggested selection of topics. The references are provided chapter by chapter. A computer disk is sup-plied with the text. This contains Minitab macros for analyses described in Chapters 2, 6, 9, 10, 11, 12 and 13; a listing of the contents of the disk appears in Appendix A and other comments are made in the appropriate chapters. Ryan has succeeded very well in his aim. The text is very readable and up-to-date, and will find a place on the shelves of research scientists as well as in the classroom.

Contents:
1. Practical reasons for using robust methods
2. A foundation for robust methods
3. Estimating measures of location and scale
4. Confidence intervals in the one-sample case
5. Comparing two groups
6. One-way and higher designs
7. Correlation and related issues
8. Robust regression
9. More regression methods

This excellent book gives a nice treatment of modern robust methods with special attention to their practical application to data. It is explained in the book how to download with one command all the needed S-Plus functions for the applications. It is of course well known that the classical methods described in the introductory courses, with their assumptions on normality, equal variance, ... can be very unsatisfactory in practice. The theory of robustness of Huber and Hampel in the sixties was a major starting point for dealing with these and other practical problems. Also the more recent resampling methods and the growing power of computers led to new solutions. The present book starts with a relatively non-technical introduction to the theory of robust methods. Chapters 4 through 9 form the main part of the book, giving clear descriptions and practical advice for the methods to use for confidence intervals, hypothesis tests, regression problems, etc. Each of these chapters ends with a number of exercises, which makes the book interesting for using it as a course text.

Contents:
1. Probability spaces
2. Integration
3. Independence and product measures
4. Convergence of random variables and measurable functions
5. Conditional expectation and an introduction to martingales
6. An introduction to weak convergence

Readership: Graduate students and researchers requiring measure theory or probability

This text provides ample and appropriate material for a one-semester introduction to measure theory for probability and statistics students among others. The book is very well written. Its stated main goal of the core material in the first four chapters is to provide a "fast track, self-contained approach to martingale theory". To impart intuition via written-words is difficult, but this author succeeds exception-ally well in several places. The discussion of the linkage between uniform integrability and truncation is a case in point. The reader might note that conditional expectation is presented using Hilbert space and projection ideas, rather than by means of the Radon-Nikodym theory. Also, the weak convergence in the last chapter is of sequences of real-valued random variables.

Contents:
1. Introduction
2. Markov and semi-Markov models as a basis for the mathematical analysis of system reliability
3. Methods for investigating homogeneous and non- homogeneous point processes (event flows)
4. Fault trees C the current state of research
5. Theory of redundant systems
6. Monte Carlo methods
7. Reliability analysis using perturbation methods
8. Stiff processes in reliability analysis
9. Variance reduction methods
10. Analytical-statistical methods for rapid simulation of repairable systems with structure redundancy
11. Measures of reliability importance of components

Drs. Kovalenko and Kuznetsov from Kyiv, Ukraïna (sic) are distinguished experts in reliability. Kovalenko in particular linked with the late Boris Gnedenko. In this book they collaborate with Dr. Pegg on a review of the mathematical theory, relating it to the practical requirements of applied researchers and engineers. Statistical inference they leave to others.
There is a compression which can lead to a definition that relies on undefined concepts deeper than that being expounded. This alone would make the book difficult to use as a stand-alone text. Room might have been found for further development, proofs and stronger examples. Notation is not standardized between chapters, nor is the content developed in a pedagogic fashion. Each chapter is given its own references, many of which are unpublished and old. As an example, there are seventy-nine items in Chapter 4 on "the current state of research". Of these twenty-five are unpublished and sixty-four are more than ten years old. Chapter 3 is introduced with "For recent developments ...see...(1982)." The main sources for the Monte Carlo methods date from 1964 and 1967.
Much of the above criticism could have been forestalled had the authors read each other's contributions critically. The book does not even appear to have been thoroughly proof-read. Ironically, what could easily have been a major contribution to reliability theory is marred by an editorial failure to attend to the quality of the product.

Contents;
PART I : Sources of Recursive Methods
1. Traditional problems
2. Rate of convergence
3. Current problems
PART II : Linear Models
4. Causality and excitation
5. Linear identification and tracking
PART III: Nonlinear Models
6. Stability
7. Nonlinear identification and control
PART IV : Markov Models
8. Recurrence
9. Learning

Readership: Statisticians, probabilists, engineers with advanced mathematical background

Interactive methods in parameter estimation, prediction problems, control and adaptive control, learning algorithms, tracking, etc., have become a popular subject since computerization made its first steps. Historically these problems were frequently studied independently in various areas of mathematics and engineering sciences. The book contributes remark-ably to the unified approach bridging various results and revealing the common elements between seemingly different statistical techniques. The presentation of the material is based on the martingale theory and the mathematical background of a reader must be rather high. Short discussions of literature sources placed at the end of every chapter are very informative while still concise. This book will be an aid to researchers familiar with the fundamentals of probability and statistics.

Contents:
PART 1. Probability Spaces, Random Variables, and Expectations
PART 2. Independence and Sums
PART 3. Convergence in Distribution
PART 4. Conditioning
PART 5. Random Sequences
PART 6. Stochastic Processes

This very comprehensive text aims to give graduate students an up-to-date introduction to a wide variety of topics. Measure theory is not a prerequisite but forms part of the main development. A feature is the number of problems (over 1200), included in the general text, and often containing proofs of the theory; hints or solutions to around 300 of the problems are understood to be available on a Web site. Parts 1 and 2 contain mainly standard topics, including random walk, laws of large numbers, and characteristic functions, as well as some of the development of measure and integration theory. Part 3 has the machinery available to include infinitely divisible and stable distributions, as well as the invariance principle and Brownian motion. Part 5 includes martingale, renewal, time-homogeneous Markov, exchangeable and stationary sequences. Part 6 introduces point processes, Lévy processes, Markov processes, interacting particle systems, and diffusion and stochastic calculus. The appendices include an extensive bibliography with directions to specialist areas.

Contents:
1. Introduction
2. Framework
3. Applications to stochastic ordinary differential equations
4. Stochastic partial differential equations

Readership: Probabilists, theoretical physicists, engineers with strong mathematical background

In modelling systems in physics, engineering and economics, one often encounters situations where the parameters governing the evolution of the system are only partially known or subject to some random "noise". To account for this lack of information one introduces stochastic noise to the model which results in stochastic ordinary or partial differential equations. In this book the authors give a comprehensive introduction to stochastic partial differential equations. Their approach is based on white noise analysis and on the interpretation of solutions to stochastic partial differential equations as functions from the space of parameters into a suitable space of stochastic distributions. First the necessary mathematical tools (mainly from functional analysis) for the analysis are provided and applications to ordinary stochastic differential equations are given. The main emphasis of the book is on stochastic partial differential equations. Examples discussed include the stochastic Poisson equation, the stochastic transport equation and certain non-linear stochastic partial differential equations as the stochastic pressure equation. While some of the material in the book is quite abstract-inevitably given the complexity of the subject-the authors give motivations and "real world" examples whenever possible. Moreover, this book is quite self-contained which makes it a "must buy" for every researcher who comes across stochastic partial differential equations in his work.

Contents:
1. Formulations of large deviation theory in terms of the Laplace principle
2. First example: Sanov's theorem
3. Second example: Mogulskii's theorem
4. Representation formulas for other stochastic processes
5. Compactness and limit properties for the random walk model
6. Laplace principle for the random walk model with continuous statistics
7. Laplace principle for the random walk with discontinuous statistics
8. Laplace principle for the empirical measures of a Markov chain
9. Extensions of the Laplace principle for the empirical measures of a Markov chain
10. Laplace principle for continuous-time Markov processes with continuous statistics

The authors give a remarkable account on how the linear theory of weak convergence of probability measures can be applied to develop the non-linear theory of large deviations. One of the highlights of this scholarly-written textbook is the representation of normalized logarithms of expectations by variational formulas that have a stochastic optimal control interpretation. As a result, a wide range of problems is tackled by a unified and slick procedure. An additional feature of this up-to-date and innovative book is that a major portion of the obtained results appear here for the first time.

Contents:
1. Introduction
2. Graphical solution of linear programming
3. Introduction to matrix construction
4. Interpreting main output
5. Examples and exercises
6. Negative coefficients and negative factor levels
7. Special techniques
8. More examples and exercises
9. Interpreting range analysis output
10. Some complications
11. Debugging your model
12. Sensitivity analysis
13. Representing risk and uncertainty
14. Further practical issues

I strongly recommend this text. It is an exceptional book: there are no symbols, no simplex method, no modelling languages nor matrix generators, just examples described in terms of words, numbers and tables giving the matrix form of the problems. How many linear programming texts have a chapter on debugging a model? The author is completely unfazed by complexities of non-linear or stochastic programming. Using simplified settings, he shows how to formulate these complex problems as linear programs. Novices will learn the wide range of circumstances that can be modelled and experienced linear programmers will learn how to extend their models to incorporate real-world complexities. If you have an interest in solving mathematical programs, whether large or small, this text will be of practical value to you.

Contents:
1. Introduction
2. Estimation
3. Testing hypotheses
4. One-way ANOVA
5. Multiple comparison techniques
6. Regression analysis
7. Multifactor analysis of variance
8. Experimental design
9. Analysis of covariance
10. Estimation and testing in general Gauss-Markov models
11. Split plot models
12. Mixed models and variance components
13. Checking assumptions, residuals, and influential observations
14. Variable selection and collinearity

This revision of the original 1987 book [Short Book Reviews, Vol. 8, p.3] retains its fairly mathematical character and has gone from 380 to 452 pages, a nineteen percent increase. Log-linear models, the former Chapter 15, and the former "Appendix F: Approximate methods for unbalanced ANOVAs" are gone, and there is more on "ANOVA-type" models and a new Appendix F. The writing style is inviting and the text often ex-presses the author's feelings in friendly and affable ways. The computing aspects of regression are de-emphasized and the text leans more towards well- prepared statistics students. If you liked the first edition, I did, you will be pleased by the second. There are one hundred and twenty-six references.

Contents:
1. Introduction
2. Background concepts and definitions
3. Shape spaces
4. Some stochastic geometry
5. Distributions of random shapes
6. Some examples of shape analysis

Readership: Statisticians, probabilists and experimental scientists with some background in differential geometry

The 'shape' of an object is what remains after all information regarding size, location and orientation is discarded. For example, all squares share the same shape. In the early 1980s, a formal theory of shape was developed, albeit from different perspectives, by D.G. Kendall and F. Bookstein. More recently, the subject has found applications in areas as diverse as archaeology, biology, medicine and vision. Although some excellent reviews exist, a textbook solely devoted to shape analysis seems to be missing. The present monograph will be a useful introduction for anyone interested in shape, as well as a helpful source of reference for researchers in the field.

Contents:
1. Introduction
2. Paired samples
3. Independent samples
4. Several independent samples
5. Simple regression
6. Overview
APPENDIX A: Geometric Tool Kit
APPENDIX B: Statistical Tool Kit
APPENDIX C: Computing
APPENDIX D: Alternative Test Statistic
APPENDIX E: Solutions to Exercises

At the heart of the linear model is the Pythagorean triangle, which decomposes the data into a component corresponding to a fitted regression and a component corresponding to error. Despite the simplicity of this idea, we all know that students have a hard time understanding the geometric interpretation of basic linear methods. However, the combined tools of projection and orthogonal decomposition are essential ingredients in an undergraduate understanding of statistics. Much of our discipline is concerned with the extraction of signal from noise, where projection is the tool for the job. However, the task of explaining the geometry of the linear model is not easy. Realistic sets of data live in high dimensions that cannot be graphed on a page. Nevertheless, the authors of this book have done an excellent job of explaining clearly what needs to be said. I would particularly recommend it as a supplementary text in an under-graduate course. Among the mass-market statistics cook-books, it makes a refreshing change.

Contents:
1. Basic remarks on multivariate probability distributions
2. Elements of the theory of point statistical estimation in the multivariate case
3. Techniques for constructing unbiased estimators
4. Applications of unbiased estimators
APPENDIX 1: Tables of Unbiased Estimators
APPENDIX 2: On Evaluating Some Multivariate Integrals
APPENDIX 3: Partitions and Some Multivariate Statistical Problems

Those working on estimation problems involving multivariate distributions will find this book a useful addition to their library. It follows on from Unbiased Estimators and Their Applications. Volume 1: Univariate Case. However, the work is largely self-contained. Certain techniques that are straightforward extensions of the univariate case have been dealt with in Volume 1 and are therefore not covered here; to that extent it is not comprehensive. There are also a number of references to Volume 1 and a thorough reading would involve having that volume available. The first two chapters give a fairly condensed account of multivariate point estimation theory. Chapter 3 considers (parametric) multivariate density estimation and techniques for determining minimum variance unbiased estimators. The applications quoted in the final chapter tend to be quite technical. In some cases considerable effort was involved in setting up the context of the application which contributed to a sense of anticlimax when the application finally appeared. In Appendix 1, the tables of unbiased estimators occupy only thirty-one pages, eleven being devoted to the multivariate normal, in contrast to the one hundred and ninety-one pages of unbiased estimators in Volume 1. Presumably this reflects the difficulty of constructing tractable non-trivial multivariate extensions of the common univariate distributions. Typographical errors are not uncommon and occasionally the English translation adheres too closely to the original Russian grammatical constructions; however the meaning is mostly clear.

Contents:
0. Preliminaries
1. An introduction to codes
2. Efficient encoding
3. Noiseless coding
4. The main coding theory problem
5. Linear codes
6. Some special codes
7. An introduction to cyclic codes

The book is written for undergraduate-level students and mixes some elementary information theory, in the Shannon sense, with a somewhat more substantial introduction to algebraic coding for the purpose of error correction and detection. The section on information theory introduces entropy and derives the coding theorem for noiseless channels without memory. The derivation is done using variable-length codes and the Kraft inequality. Huffman coding for binary code alpha-bets is described and its optimality is proved. There is no mention of the asymptotic equipartition property. The coding theorem for a noisy binary symmetric channel is stated but not proved. The section on coding theory describes discrete noisy communication channels and then discusses the general idea of error correcting and detecting codes. Linear codes are defined and some properties are developed. Hamming, Reed-Muller, Golay, and cyclic codes are described in varying amounts of detail. Sphere-packing, Plotkin, and Gilbert-Varshamov bounds on performance are derived. Exercises are provided at the end of sections to help the reader.

Contents:
1. Introduction
2. Least squares and methods of moments
3. Extremum estimators and method-of-moments estimators
4. Maximum likelihood estimation
5. Parametric estimators for multiple equations
6. Nonlinear models and generalized method of moments
7. Nonparametric density estimation
8. Nonparametric regression
9. Semi-parametrics
10. Semi-nonparametrics
APPENDIX: Gauss Programs for Selected Topics

The purpose of the book is to discuss recent developments in econometric theory on methods of mo-ments and semi-parametric econometric methods for limited-dependent-variable models. Gauss programs for some of the key methods are given. The mathematical material is rigorous, and the book should be very use-ful to the intended readership, although less useful to practitioners.

Contents:
PART I : Models
1. Model building
2. Exponential family of probability distributions
PART II : Inference
3. Likelihood
4. Goodness of fit
PART III: Approximations
5. Asymptotics
6. Factoring the likelihood function
PART IV : Decisions
7. Frequentist decision-making
8. Bayesian decision-making
PART V : Examples
9. Poisson regression
10. Binomial regression
APPENDIX A: Elements of Measure Theory
APPENDIX B: Review of Probability Theory
APPENDIX C: Normal Distribution Statistics
APPENDIX D: Numerical Methods

Readership: Advanced undergraduates and graduates in statistics, statisticians, both theoretical and applied

This is an important book! Based firmly on the likelihood, here is a text that encapsulates the modern view of statistical inference and decision making from both the frequentist and Bayesian viewpoints. Rather than highlighting differences between these viewpoints, the author shows how they both have a role to play in inference. The central idea is that of drawing inferences from empirical data by means of a statistical model. For the most part the data are assumed to have been generated by a distribution in the exponential family, but the transformation and stable distributions are also considered. The text is remarkable for its clarity of exposition, the numerous examples, excellent cross references and its many insightful comments. The mathematical level is moderate, with brief reviews of the underlying measure theory and probability theory given in appendices. This is a book on theoretical statistics, and yet one is never far from the data. Throughout data is used to illustrate the theoretical results, both in the text and in the numerous exercises that accompany each chapter. Part V is devoted to two examples, a Poisson and a binomial regression which show in detail how theory is applied at each stage of an analysis. This book is highly recommended both as a text for advanced undergraduates and to any professional statistician who would like a refresher course in modern statistical theory.

Contents:
1. Populations, measurements and parameters
2. Expectations and Daniell integrals
3. Random variables and measureable functions
4. Construction of Daniell integrals
5. Least squares
6. Independence and dependence
7. Quantiles
8. Moments

This is the first volume in a two-volume work evidently designed to provide a rigorous mathematical treatment of the definitions of population characteristics including measures of location and size. "The description of populations" describes the motivation for mathematical methodology developed in this text, not the contents of the text. There are two populations discussed, the population of the United States and the number of newspapers in the United States from 1920 to 1970. Otherwise the book uses Daniell integrals to visit what might otherwise be thought of as familiar ground: Lebesgue integrals, least squares, independence, quantiles and moments. This volume is not for the mathematically faint-of-heart.

Contents:
1. Introduction
2. Looking at data
3. Formal tests of uniformity
4. Comparing populations of judges
5. Overview of models
6. Distance-based models
7. Babington-Smith, phi-models and inversions
8. Plackett-Luce, logistic and vase models
9. Marginal and ANOVA-type loglinear models
10. Latent class and unfolding models
11. Tied, partial and incomplete rankings

This scholarly book brings together a vast literature on methods for analyzing and modeling rank data. All aspects of the analysis are considered: graphical displays, comparing populations, modeling the data and handling incomplete rankings. It is a mathematical statistics book in the best sense of the word. The assumptions underlying each technique are clearly stated and results are developed mathematically. Further details and extensions are given in the exercises that accompany each chapter. Particularly helpful is the careful notational distinction the author makes between the ranks assigned to a set of objects, y, and their ordering, w. Methods are compared and contrasted, links between them and equivalences are clearly stated. However, the author does not stop at that. He carefully works through well-chosen examples, showing the readers how to think about the data and warning them of any pitfalls on the way. Some of the examples are extremely simple but illustrate precisely what is meant; the rest are drawn from such diverse fields of application as voting behaviour, feeding preferences of cows and the literary criticism. This book is suitable for a graduate course in ranking methods. It is likely to become a standard reference in this field. A welcome supplement to this volume would be a web site giving the data used in the examples and information on availability of software to implement the methods.

Contents:
1. Introduction
2. Simple univariate density estimation
3. Smoother univariate density estimation
4. Multivariate density estimation
5. Nonparametric regression
6. Smoothing ordered categorical data
7. Further applications of smoothing
APPENDIX A: Descriptions of the Data Sets
APPENDIX B: More on Computational Issues

This is an attractive book surveying clearly many of the uses of smoothing methods in statistics. The emphasis is on the application of a variety of techniques to real problems, though the common structure in the different methods is discussed carefully. The book should appeal to statisticians wishing to study the methods themselves as much as to the data analyst. The coverage in the book is broad and it will serve many useful purposes: as a guide for practitioners, as an up-to-date drawing together of widely dispersed material of interest to researchers and as a textbook suitable for advanced students. The presentation and arrangement of the material is good. Each chapter is supplemented with background material and discussion of computational issues. The book contains a very extensive reference list. Though it seems inevitable that a book such as this, in a rapidly evolving field, will date quickly, in the meantime I would strongly recommend it.

Contents:
1. Introduction
2. Modeling concepts and methods
3. The smoothness priors concept
4. Scalar least squares modeling
5. Linear Gaussian state space modeling
6. General state space modeling
7. Applications of linear Gaussian state space modeling
8. Modeling trends
9. Seasonal adjustment
10. Estimation of time varying variance
11. Modeling scalar nonstationary covariance time series
12. Modeling multivariate nonstationary covariance time series
13. Modeling inhomogeneous discrete processes
14. Quasi-periodic process modeling
15. Nonlinear smoothing
16. Other applications

The modeling of time-series data has taken many forms: frequency-domain analysis, time-domain analysis and Kalman filtering. This text treats the latter form and extends it bringing in the concept of smoothness priors. The authors have done an excellent job of presenting the theoretical underpinnings of Kalman filtering, both in the linear-Gaussian setting and the non-Gaussian nonlinear state-space setting. A strength of the text is the authors' concern with identifying the criterion for model identification and subsequent estimation. The second half of the text is devoted to many examples which explore the methods discussed in the first six chapters. Their treatment of problems of missing data and irregularly-spaced observations is also well done. The authors suggest several numerical techniques to analyze the data and cite numerous papers that will benefit researchers and analysts. A well-written and informative text on a stimulating topic.

Contents:
1. Bootstrap variable-selection and confidence sets, by R. Beran
2. Robust estimation in the logistic regression model, by A.M. Bianco and V.J. Yohai
3. The m out of n bootstrap goodness of fit tests with double censored data, by P.J. Bickel and J.-J. Ren
4. What criterion for a power algorithm? by A. Buja
5. Nonparametric estimation of global functionals of conditional quantiles, by P. Chaudhuri, K. Doksum and A. Samarov
6. High breakdown point estimators in logistic regression, by A. Christmann
7. The Guinea pig of multiple regression, by Y. Dodge
8. When is a p-value a good measure of evidence? by M.B. Dollinger, E. Kulinskaya and R.G. Standte
9. Geometric properties of pricipal curves in the plane, by T. Duchamp and W. Stuetzle
10. On robust estimation of variograms in geostatistics, by R. Dutter

11. Advances in nonparametric function estimation, by T. Gasser
12. On the philosophical foundations of statistics: Bridges to Huber's work and recent results, by F. Hampel
13. Data based prototyping, by T.M. Huber and M. Nagel
14. Robust regression with a categorical covariable, by M. Hubert and P.J. Rosseeuw
15. Robustness in discriminant analysis, by Y.S. Kharin
16. M-estimation and spatial quantiles, by V. Koltchinskii
17. Robust statistical methods in interlaboratory analytical studies, by P. Lischor
18. Estimating distributions with a fixed number of modes, by M.B. Mächler
19. Implementing M-estimation of the gamma distribution, by A. Marazzi and C. Fullieux
20. Constrained M-estimation for regression, by B. Mendes and D.E. Tyler
21. Inference for the direction of the larger of two eigenvalues: The case of circular elongation, by S. Morgenthaler and J.W. Tukey
22. High breakdown point designs, by C.H. Müller
23. On Bayesian robustness: An asymptotic approach, by D. Peña and R.H. Zamar
24. Computational aspects of trimmed single-link clustering, by E. Tabakis
25. Interval probability on finite sample spaces, by K. Weichselberger
26. On consistency of recursive multivariate M-estimators in linear models, by Y. Wu

These twenty-six papers were associated with a workshop in June, 1994, organized about the areas in statistics that Peter Huber has influenced. Most of the papers deal with issues related to robustness. The treatments, however, vary considerably from the deeply philosophical of F. Hample's "On the Philosophical Foundations of Statistics" to the specific problems associated with particular situations. Y. Dodge presents a meta-analysis of sixty re-analyses of the stack-loss data of Brownlee.
The papers were presented in honour of a distinguished statistician. Many contain interesting references to Huber's interactions with their authors. Thus, in addition to presenting valuable contributions to the field of robustness this book sheds some light on the nature and extent of Huber's influence.

Contents:
1. Nonlinear regression model and parameter estimation
2. Accuracy of estimators, confidence intervals and tests
3. Variance estimation
4. Diagnostics of model misspecification
5. Calibration and prediction

The book may be considered as a short introduction to the nonlinear regression analysis complemented by a large number of examples with an extensive computational component. All computations are based on the software, which is developed by the authors and which is an extension of the popular S-PLUS. As the authors state themselves, their intention is to prepare a `cook-book`. Perhaps, they achieved this goal: the book contains a number of well-written recipes which can be applied to nonlinear regression with additive random-error form. The regression function is assumed to be given in a closed form. As it frequently happens with cook books you can easily find a rather detailed recipe but it is much more difficult to conclude from the text why it is good or better than other recipes.

Contents:
1. Introduction
2. Frame constructions
3. Constructions for resolvable designs
4. Existence results

This book is aimed at a mathematical reader-ship. It may also be useful as a reference to statistical researchers in combinatorial experimental design. It deals mainly with the existence and construction of resolvable and near resolvable balanced incomplete block designs. Interrelationships between combinatorial structures is a key feature, especially in the use of frames for design construction purposes. A frame is a certain type of group divisible design. The reader should be aware that the definition of the latter differs from the standard statistical definition. Block sizes are not necessarily equal and no treatments in the same group appear together in a block. The book will not appeal to many statisticians due to the lack of exposition in all but the introductory sections. However, the last chapter, which gives a comprehensive table detailing the existence of resolvable balanced incomplete block designs, together with references for their construction, is likely to be useful.

Contents:
1. Introduction and preliminaries
2. Fixed-population sampling theory
3. Stochastic population sampling theory
4. Adaptive cluster sampling
5. Efficiency and sample size issues
6. Adaptive cluster sampling based on order statistics
7. Adaptive allocation in stratified sampling
8. Multivariate aspects of adaptive sampling
9. Detectability in adaptive sampling
10. Optimal sampling strategies

This monograph considers the theory and methods of adaptive sampling. Such sampling methods are particularly appropriate when information is sought on rare events that tend to cluster in the population, such as a rare animal or plant species, and refer to situations where the sampling design depends on the value of a response variable observed during the sampling process. In its simplest form, an adaptive cluster sampling design is based on an initial random sample selected from the population. For each unit in the population, a neighbourhood is defined and whenever the response value of a selected unit in the sample satisfies a specified condition, all units in its neighbour-hood are added to the sample. If any of the added units also satisfy the condition, their neighbourhoods are included as well. In this way, a network of units is created as the final sample. This concept of a network is used to develop appropriate estimators and their properties using a variety of initial sampling plans, specific conditions and neighbourhood definitions.
The explanations of the methods are concise and clear and each chapter is illustrated with practical examples. Much of the material included is based on recent research of the authors and should be useful to anyone involved in sampling from populations where conventional sampling plans are inefficient and wasteful. The gain in efficiency of these adaptive methods over conventional sampling can be considerable, particularly when the individuals of interest occur in the sparsely distributed clusters.
Adaptations of the basic ideas are extended in later chapters to stratified sampling and to the use of multivariate responses to identify the units whose neighbourhoods are to be included in the sample. The estimation formulae derived in earlier chapters are extended to the situation where the probability of detecting an individual of interest is less than one, even if it is in the sampled unit. In the final chapter the authors consider optimal sampling, strategies for fixed populations with varying degrees of model assumptions.

Contents:
PART I : Basic Probability
1. Modes of convergence
2. Partial converses to Theorem 1
3. Convergence in law
4. Laws of large numbers
5. Central limit theorems
PART II : Basic Statistical Large Sample Theory
6. Slutsky theorems
7. Functions of the sample moments
8. The sample correlation coefficient
9. Pearson's chi-square
10. Asymptotic power of the Pearson chi-square test
PART III: Special Topics
11. Stationary m-dependent sequences
12. Some rank statistics
13. Asymptotic distribution of sample quantiles
14. Asymptotic theory of extreme order statistics
15. Asymptotic joint distributions of extrema
PART IV : Efficient Estimation and Testing
16. A uniform strong law of large numbers
17. Strong consistency of maximum-likelihood estimates
18. Asymptotic normality of the maximum-likelihood estimate
19. The Cramér-Rao lower bound
20. Asymptotic efficiency
21. Asymptotic normality of posterior distributions
22. Asymptotic distribution of the likelihood ratio test statistic
23. Minimum chi-square estimates
24. General chi-square tests

This is a very useful book. It provides teachers and students with a selection of basic theorems, the proofs of which are usually "postponed" in the undergraduate courses and "assumed to be given before" in the graduate courses. With his long and valuable experience, the author has made a choice of twenty-four topics (and twenty-four theorems with proofs) in the area of asymptotic statistics, mainly consistency and central limit results; see the table of contents. Most of the proofs are given in a multivariate setting. Another useful aspect of the book is the good number of exercises after each of the twenty-four sections and the more than sixty pages devoted to their solutions. Also interesting for teaching is that further exercises can be found an a web page for the course. The book is ideal for self study and all instructors of statistical inference will improve their teaching by consulting it.

Contents:
1. Introduction
2. Moving average and autoregressive processes
3. Introduction to Fourier analysis
4. Spectral theory and filtering
5. Some large sample theory
6. Estimation of the mean and autocorrelations
7. The periodogram, estimated spectrum
8. Parameter estimation
9. Regression, trend, and seasonality
10. Unit root and explosive time series

This is a thoroughly revised version of the first edition originally published in 1976. References in this new edition are as recent as 1995. The book leans heavily towards time series methodology of interest in econometrics; indeed Chapter 10 now incorporates tests for unit roots in univariate and vector pro-cesses. Other additional material incorporated in this edition includes partial autocorrelations, long memory processes, the Kalman filter, central limit theorems for martingale differences, and an expanded treatment of nonlinear estimation. The strength of the book is its comprehensive coverage of important theoretical concepts and results which makes it very useful, if not indispensable, for research purposes. The theorem-proof style will not render it appealing to practitioners, and indeed there is relatively little discussion of computational tools. There is a very nice list of principal results, but the index is insufficient for a book of this size and quality. Overall though, this is a very useful book for researchers, and a must for all library statistical book collections.

Contents:
1. Introduction
2. Gambling houses and the conservation of fairness
3. Leavable gambling problems
4. Nonleavable gambling problems
5. Stationary families of strategies
6. Approximation theorems
7. Stochastic games

This monograph is in the tradition of the now classical How To Gamble If You Must: Inequalities For Stochastic Processes by L.E. Dubins and L.J. Savage published in 1965, in which a general theory of discrete time stochastic control was developed, using the vehicle of gambling terminology and examples. It provides an introduction to these ideas, and also to more recent developments in gambling theory. The conventional assumption of countably additive probability measures is made and state spaces are usually taken to be countable. There are problem sets at the end of each chapter. Prerequisites are, essentially, a course in measure theoretic probability; no prior knowledge of gambling theory, game theory or stochastic control is necessary.

Contents:
1. Brownian motion I
2. Stochastic integration
3. Brownian motion II
4. Partial differential equations
5. Stochastic differential equations
6. One-dimensional diffusions
7. Diffusion, as Markov processes
8. Weak convergence

The book summarizes the author's experience in teaching many sophisticated aspects of stochastic calculus, and correspondingly it looks very much like an extended and carefully edited version of lecture notes. The coverage of the topics is extensive, challenging for the reader, and made on a high professional level. The statement, which is posted on the back cover page and says that this single volume is "saving you time, effort, and expense", does not fairly reflect the reality. First, you have to be a very devoted student to pass through the whole book and it would not be a short-term effortless exercise at all. Secondly, the author's calculations made in the preface about a savings of possibly "more than $1000", if you buy his book rather than the 18 books which he kindly cites and assumedly summarizes, sounds very much like "buy one and get 18 free." As in all other similar deals there is a catch: after reading a few sections you notice that the intensity of referencing to the previously published book by the same author (Durrett 1995) is so high that you have no choice but to buy that book. The list of references provided is very parsimonious, and may be usefully extended by references to the books which were authored by such major players in the topic as Feller, Girsanov, Khasminskii, Kolmogorov, Prokhorov, and Scorokhod, whose results are so frequently cited by the author.

Contents:
1. Measure theory
2. Integration
3. Fourier analysis
APPENDIX A: Metric Spaces
APPENDIX B: On Lp Matters
APPENDIX C: A Non-Measurable Set of the Interval [0,1]

Readership: Students studying analysis (measure theory) or probability or mathematical statistics

The first edition of this book was published in 1986 [Short Book Reviews, Vol. 6. p.27]. The second edition differs from the first one because misprints are corrected, but the set-up remains the same. The presentation is nice. The book is well suited as a textbook in an introductory course. The authors also give suggestions to collateral reading so that a student can widen his knowledge. Each section ends with a list of exercises, not all of them being trivial. The first two sections of Chapter 1 introduce, in a heuristic but mathematically acceptable way, concepts of probability. These sections serve as a motivation to start with measure theory. As the title suggests, more advanced topics in probability are not treated.

Contents:
PART I : Integration Relative to Daniell Measures
1. Reisz spaces
2. Measures on semirings
3. Integrable and measurable functions
4. Lebesgue measure on R
5. Lp spaces
6. Integrable functions for measures on semirings
7. Radon measures
8. Regularity
PART II : Operations on Measures and Product Measures
9. Induced measures and product measures
10. Radon-Nikodym derivatives
11. Images of measures
12. Change of variables
13. Stieltjes integral
14. The Fourier transform in Rk
PART III: Convergence of Random Variables; Conditional Expectation
15. The strong law of large numbers
16. The central limit theorem
17. Order statistics
18. Conditional probability
PART IV : Operations on Radon measures
19. ì-Adequate family of measures
20. Radon measues defined by densities
21. Images of Radon measures and product measures
22. Operations on regular measures
23. Haar measures
24. Convolution of measures

This book provides a careful and thorough coverage of measure and integration theories, and would be appropriate for a graduate-level advanced text or refrence source. A few problems are included at the end of the chapters; most of these are lengthy and challenging multipart problems. For example, Stirling's approximation is a six-part full page exercise in Chapter 12. The probability portion is brief. It serves only as a reasonable illustration of a couple of im-\portant areas in probability, not as a text for a probability course.

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

Reviewer:
Institute	Queen's University
Place	Kingston, Canada
Name	J.H. Davis

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

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

Reviewer:
Institute	University of Chicago
Place	Chicago, U.S.A.
Name	S.M. Stigler

Title	BONFERRONI-TYPE INEQUALITIES WITH APPLICATIONS.
Author	J. Galambos. and I. Simonelli.
Publisher	New York: Springer-Verlag, 1996, pp. ix + 269, US$54.95.

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

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

Reviewer:
Institute	Calamai Associates
Place	Ottawa, Canada
Name	Peter Calamai

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

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

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

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

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

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

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

Reviewer:
Institute	Simon Fraser University
Place	Burnaby, Canada
Name	M.A. Stephens

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

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

Reviewer:
Institute	Oak Ridge National Laboratory
Place	Oak Ridge, U.S.A.
Name	V.V. Fedorov

Reviewer:
Institute	University of Kent
Place	Canterbury, U.K.
Name	H. Tong

Reviewer:
Institute	Surrey University
Place	Guildford, U.K.
Name	M.J. Crowder

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

Reviewer:
Institute	Harvard School of Public Health
Place	Boston, U.S.A.
Name	J.G. Ibrahim

Reviewer:
Institute	Duke University
Place	Durham, U.S.A.
Name	R.L. Wolpert

Reviewer:
Institute	ETH Zürich,
Place	Zürich, Switzerland
Name	R. Frey

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

Reviewer:
Institute	ETH Zürich
Place	Zürich, Switzerland
Name	F. Delbaen

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

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

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

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

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

Reviewer:
Institute	University of Warwick
Place	Coventry, U.K.
Name	M.C. van Lieshout

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