Short Book Reviews On-Line 1998

THE EM ALGORITHM AND EXTENSIONS. G.J. McLachlan and	T. Krishnan.
PROBABILITY, STOCHASTIC PROCESSES, AND QUEUEING THEORY.	R. Nelson.
SOME ASPECTS OF BROWNIAN MOTION, PART II: SOME RECENT MARTINGALE PROBLEMS.	M. Yor.
ANALYSIS OF VARIANCE IN STATISTICAL IMAGE PROCESSING.	L. Kurz and M.H. Benteftifa.
FUNCTIONAL DATA ANALYSIS.	J.O. Ramsey and B.W. Silverman.
INTERIOR POINT APPROACH. THEORY AND ALGORITHMS FOR LINEAR OPTIMIZATION: …	C. Roos, T. Terlaky, and J-.Ph. Vial.
LEADING PERSONALITIES IN STATISTICAL SCIENCES. From the Seventeenth Century to the Present.	N.L. Johnson and S. Kotz (Eds).
PROVE IT WITH FIGURES.	H. Zeisel and D. Kaye.
MAD COWS AND MOTHER'S MILK: THE PERILS OF POOR RISK COMMUNICATION.	D. Powell and W. Leiss.
ENCYCLOPEDIA OF BIOSTATISTICS. Volumes 1 (A-Cox), 2 (Cra-G), 3 (H-Mea), 4 (Med-Pre), 5 (Pri-Sph), 6 (Spi-Z/Index).	P. Armitage and T. Colton(Eds-in-Chief).
VISUAL REVELATIONS: Graphical Tales of Fate and Deception from Napoleon Bonaparte to Ross Perot.	H. Wainer.
DÉMOGRAPHIE. APPROCHE STATISTIQUE ET DYNAMIQUE DES POPULATIONS.	H. Leridon et L. Toulemon.
MATHEMATICAL AND STATISTICAL METHODS FOR GENETIC ANALYSIS.	K. Lange.
APPLIED SURVIVAL ANALYSIS.	C.T. Le.
CLINICAL TRIALS IN ONCOLOGY.	S. Green, J. Benedetti and J. Crowley.
MANAGEMENT OF DATA IN CLINICAL TRIALS.	E. McFadden.
SETTING ENVIRONMENTAL STANDARDS: The Statistical Approach to Handling Uncertainty and Variation.	V. Barnett and A. O'Hagan.
HANDBOOK OF EXPERIMENTAL METHODS FOR PROCESS IMPROVEMENT.	D. Drain.
IMPROVING EFFICIENCY BY SHRINKAGE. The James-Stein and Ridge Regression Estimators.	M.H.J. Gruber.
APPLIED SMOOTHING TECHNIQUES FOR DATA ANALYSIS. THE KERNEL APPROACH WITH S-PLUS ILLUSTRATIONS.	A.W. Bowman and A. Azzalini.
THEORY OF SAMPLE SURVEYS.	M.E. Thompson.
BAYESIAN METHODS FOR FINITE POPULATION SAMPLING.	M. Ghosh and G. Meeden.
DATA DRIVEN STATISTICAL MODELS.	P. Sprent.
BOOTSTRAP METHODS AND THEIR APPLICATIONS.	A.C. Davison and D.V. Hinkley.
MARKOV CHAIN MONTE CARLO: Stochastic Simulation for Bayesian Inference	D. Gamerman.
THE ANALYSIS OF PROXIMITY DATA.	B.S. Everitt and S. Rabe-Hesketh.
S+SPATIALSTATS: USER'S MANUAL FOR WINDOWS AND UNIX.	S.P. Kaluzny, S.C. Vega, T.P. Cardoso and A.A. Shelly.
PETIT COURSE DE STATISTIQUE.	K. Krickeberg.
THE THEORY OF DISPERSION MODELS.	B. Jørgensen.
STATISTICAL ANALYSIS OF EXTREME VALUES. FROM INSURANCE, FINANCE, HYDROLOGY AND OTHER FIELDS.	R.-D. Reiss and M. Thomas.
STATISTICS IN FINANCE.	D.J. Hand and S.D. Jacka (Eds.).
LECTURES ON THE MATHEMATICS OF FINANCE.	I. Karatzas.
FINANCIAL NETWORKS: STATISTICS AND DYNAMICS.	A Nagurney and S. Siokos.
AN INTRODUCTION TO ECONOMIC THEORY. Measure-Theoretic Probability and Statistics with Applications to Economics.	A.R. Gallant.
PROBABILISTIC MODELLING.	I. Mitrani.
STATISTICAL ISSUES IN DRUG DEVELOPMENT.	S. Senn.
STATISTICS FOR ENVIRONMENTAL BIOLOGY AND TOXICOLOGY.	W.W. Piegorsch and A.J. Bailer
ASYMPTOTIC BEHAVIOUR OF LINEARLY TRANSFORMED SUMS OF RANDOM VARIABLES.	V. Buldygin and S. Solnstev.
LIMIT THEOREMS IN CHANGE-POINT ANALYSIS. M. Csörgö and	L. Horváth.
EXPERIMENTAL STOCHASTICS.	O. Moeschlin, E. Gryco, C. Pohl and F. Steinert.
FUNDAMENTAL LIMITATIONS IN FILTERING AND CONTROL.	M.M. Seron, J.H. Braslavsky and G.C. Goodwin.
STATISTICS FOR THE ENVIRONMENT 3: Pollution Assessment and Control.	V. Barnett and K.F. Turkman (Eds.).
REGIONAL FREQUENCY ANALYSIS. An Approach Based on L-Moments.	J.R.M. Hosking and J.R. Wallis.
A HISTORY OF MATHEMATICAL STATISTICS FROM 1750 TO 1930.	A. Hald
STATISTICAL VISIONS IN TIME:a History of time Series Analysis 1662 – 1938	J.L. Klein
LIFE'S OTHER SECRET.	I. Stewart
EVOLUTIONARY GAMES AND POPULATION DYNAMICS.	J. Hofbauer and K. Sigmund.
STATISTICS IN HUMAN GENETICS.	P. Sham.
PLANS D'EXPÉRIENCES: APPLICATIONS À L'ENTREPRISE.	J.-J. Droesbeke, J. Fine et G. Saporta.
CUMULATIVE SUM CHARTS AND CHARTING FOR QUALITY IMPROVEMENT.	D.M. Hawkins and D.H. Olwell
PROGRAMMING WITH DATA: A GUIDE TO THE S LANGUAGE.	J.M. Chambers.
INTODUCTION TO DESIGN AND ANALYSIS OF EXPERIMENTS.	G.W. Cobb
STATISTICAL TESTS FOR MIXED LINEAR MODELS.	A.I. Khuri, T. Mathew and B.K. Sinha
NONPARAMETRIC SMOOTHING AND LACK-OF-FIT TESTS.	J.D. Hart.
MODERN SIMULATION AND MODELING.	R.Y. Rubinstein and B. Melamed
SEMIPARAMETRIC METHODS IN ECONOMETRICS.	J.L. Horowitz
NONRESPONSE IN HOUSEHOLD INTERVIEW SURVEYS.	R.M. Groves and M.P. Couper
MATHEMATICAL TOOLS FOR APPLIED MULTIVARIATE ANALYSIS. Revised edition.	J.D. Carroll and P.E. Green. With contributions by A. Chaturvedi.
ROBUST NONPARAMETRIC STATISTICAL METHODS.	T.P. Hettmansperger and J.W. McKean
APPLIED MULTIVARIATE STATISTICS IN GEOHYDROLOGY AND RELATED SCIENCES.	C.E. Brown.
INTRODUCING MULTILEVEL MODELLING.	I. Kreft and J. de Leeuw
SMALL AREA ESTIMATION IN SURVEY SAMPLING.	P. Mukhopadhyay.
LOCAL STEREOLOGY.	E.B. Vedel Jensen.
COLLECTED WORKS OF JAROSLAV HÁJEK – WITH COMMENTARY.	Compiled by M. Hušková, R. Beran and V. Dupaè
STATISTIQUE THORIQUE ET APPLIQUEE : Tome 1. Statistique descriptive et bases de l'inférence statistique.	P. Dagneli.
CONTINUOUS -TIME MARKOV CHAINS AND APPLICATIONS. A SINGULAR PERTURBATION APPROACH.	G.G. Yin and Q. Zhang.
STOCHASTIC APPROXIMATION ALGORITHMS AND APPLICATIONS.	H.J. Kushner and G.G. Yin
OPTIMAL PORTFOLIOS. Stochastic Models for Optimal Investment and Risk Management in Continuous Time.	R. Korn
LOSS MODELS: FROM DATA TO DECISIONS.	S.A. Klugman, H.H. Panjer and G.E. Willmot. With assistance from G.G. Venter
ENGINEERING RELABILITY.	R.E. Barlow
STATISTICAL CONTROL BY MONITORING AND FEEDBACK ADJUSTMENT.	G. Box and A. Luceño
MODERN THEORY OF SUMMATION OF RANDOM VARIABLES.	V.M. Zolotarev
MARTINGALE APPROXIMATION.	Y.V. Borovskikh and V.S. Korolyuk
MODELLING EXTREMAL EVENTS FOR INSURANCE AND FINANCE.	P. Embrechts, C. Klüppelberg and T. Mikosch.
THE ECONOMETRICS OF FINANCIAL MARKETS.	J.Y. Campbell, A.W. Lo and A.C. MacKinlay.
MARTINGALE METHODS IN FINANCIAL MODELLING.	M. Musiela and M. Rutkowski.
MASS TRANSPORTATION PROBLEMS: Theory; Volume II: Applications.	S.T. Rachev and L. Rüschendorf
INTEGRAL, PROBABILITY, AND FRACTAL MEASURES.	G.A. Edgar
BAYESIAN ECONOMICS THROUGH NUMERICAL METHODS: A GUIDE TO ECONOMETRICS AND DECISION-MAKING WITH PRIOR INFORMATION.	J.H. Dorfman.
MIXED POISSON PROCESSES.	J. Grandell.
QUASI-LIKELIHOOD AND ITS APPLICATION. A General Approach to Optimal Parameter Estimation.	C.C. Heyde.
FOUNDATIONS OF MODERN PROBABILITY.	O. Kallenberg.
A FIRST COURSE IN MULTIVARIATE STATISTICS.	B. Flury.
ANALYSIS OF INCOMPLETE MULTIVARIATE DATA.	J.L. Schafer.
APPLYING GENERALIZED LINEAR MODELS.	J.K. Lindsey.
MODERN APPLIED STATISTICS WITH S-PLUS, 2nd edition.	W.N. Venables and B.D. Ripley.
EXPONENTIAL FAMILIES OF STOCHASTIC PROCESSES.	U. Kuechler and M. Soerensen.
RETRIAL QUEUES.	G.I. Fallin and J.G.C. Templeton.
SEQUENTIAL ESTIMATION.	M. Ghosh, N. Mukhopadhyay and P.K. Sen.
GRAPHICAL MODELS.	S.L. Lauritzen.
MODEL-ORIENTED DESIGN OF EXPERIMENTS.	V.V. Fedorov and P. Hackl.
ROBUST PLANNING AND ANALYSIS OF EXPERIMENTS.	C.H. Müller.
GEOMETRICAL FOUNDATIONS OF ASYMPTOTIC INFERENCE.	R.E. Kass and P.W. Vos.
STOCHASTIC ANALYSIS.	P. Malliavin.

Contents:
1. General introduction
2. Examples of the EM algorithm
3. Basic theory of the EM algorithm
4. Standard errors and speeding up convergence
5. Extensions of the EM algorithm
6. Miscellaneous topics

The EM algorithm has been around for a long time and is routinely used in incomplete data situations where there are missing, censored or grouped data, or where truncated distributions are involved. It can also be used in some complete data situations for which maximum likelihood estimation has been found intractable, by reformulating the problem as an in-complete data problem. It is surprising therefore that a text such as this has not appeared earlier. Evidently it is the first one that focuses on the algorithm and gives a comprehensive treatment of the subject. Those wishing to understand the subject from scratch should find the early chapters give a very good introduction albeit with some effort on the part of the reader. There are plenty of good motivating examples drawn from a broad specturm of contexts these serve to reinforce the wide applicability of the method. The important issues of convergence and convergence rates are well covered and the recent evolution of the method to handle problems outside the scope of the conventional EM algorithm is discussed in some detail. Although not set up as a teaching text with exercises at the end of each chapter etc., the book should help promote the teaching of this important subject in postgraduate and appropriate undergraduate courses.

Contents:
1. Introduction
2. Randomness and probability
3. Combinatorics
4. Random variables and distributions
5. Expectation and fundamental theorems
6. The Poisson process and renewal theory
7. The M/G/1 queue
8. Markov processes
9. Matrix geometric solutions
10. Queueing networks
11. Epilogue and special topics
APPENDIX A: Types of Randomness
APPENDIX B: Combinatorial Equalities and Inequalities
APPENDIX C: Tables of Laplace Transforms and Generating Functions
APPENDIX D: Limits and Order Relationships
APPENDIX E: List of Common Summations

This book is most unusual but at the same time intriguing. It is unusual, i.e. in the sense that the author gears his attention towards computer performance problems that play the role of Leidmotiv. He approaches his goal by making the subject as accessible as possible without unnecessarily compromising any one aspect of the theory. At the end of each chapter, a summary is given as well as succinct historical notes. Further-more, each chapter ends with a vast set of exercises classified according to intrinsic importance and level of difficulty. On the intriguing side we find a gentle treatment of a vast number of subjects that most often can only be found in advanced level texts. Among them semi-Markov processes, matrix geometric systems, phase distributions, queueing networks and a variety of different queueing models. Some readers might object to the lack of sophistication, but the book is remarkable in giving a lucid bird's eye view on a wealth of stochastic systems.

Contents:
10. On principal values of Brownian and Bessel local times
11. Probabilistic representations of the Riemann zeta function
12. Some examples and applications of enlargements of filtrations
13. Martingale inequalities at any time
14. On the martingales which vanish on the set of Brownian zeroes
15. On Azéma's martingales and the chaos representation property
16. The filtration of truncated Brownian motion
17. The Brownian filtration, Tsirel'son's examples, and Walsh's Brownian motions
18. Complements relative to Part I (Chapters 1 to 9)

Packed into this short volume are many deep applications involving Brownian motion and related martingales of powerful and relatively new methodologies from stochastic analysis. The staccato presentation of the many seemingly disparate topics does not diminish the reader's sense of excitement about the great potential for the illustrated technology. The author conveys well the 'hot-off-the-press' flavour of the material: Note especially the Epilogue to Chapter 17. Several papers not included in the bibliography are discussed. Updates and additional references regarding the earlier volume are also included. In total, there are over one hundred and thirty-five references, with eighty-five of these dated 1988 and later. This book should be a stimulating text and reference for specialists, and may, in the metaphor of the Foreword, make the gold miners in the Brownian hills a little better equipped for their increasingly arduous work.

Contents:
1. Introduction
2. Statistical linear models
3. Line detection
4. Edge detection
5. Object detection
6. Image segmentation
7. Radial masks in line and edge detection
8. Performance analysis
9. Some approaches to image restoration

At last we have a book that, by its effective use of classical statistical techniques in image pro-cessing, acknowledges that images are statistical data. The use of experimental design methodology in the pro-cessing of images corrupted by relatively high Gaussian noise is detailed. The text was written as a graduate course, and would require a practical application of the procedures to be fully appreciated. The authors recommend this approach and so have deliberately chosen not to include problems and exercises. In teaching their course they used software developed over the years, which is also not included.

Contents:
1. Introduction
2. Notation and techniques
3. Representing functional data as smooth functions
4. The roughness penalty approach
5. The registration and display of functional data
6. Principal components for functional data
7. Regularized principal components analysis
8. Principal components analysis of mixed data
9. Functional linear models
10. Functional linear models for scalar responses
11. Functional linear models for functional responses
12. Canonical correlation and discriminant analysis
13. Differential operators in functional analysis
14. Principal differential analysis
15. More general roughness penalties
16. Some perspectives on FDA
APPENDIX: Some Algebraic Functional Techniques

Readership: Researchers and statisticians working with repeated measures and comfortable with mathematics at a senior undergraduate level

This excellent book presents a large coherent collection of techniques for the analysis of data of repeated measurements of smooth functions. Growth curves are an example. Methods for one and higher-dimensional phenomena are discussed.
By abbreviating their impressive work to FDA, the authors invite a comparison with its temporal and lexicographical precident, namely EDA or Exploratory Data Analysis by J.W. Tukey although no reference is given. The similarities include the wealth of techniques motivated by and illustrated with data. Indeed FDA begins: Figure 1.1 provides .. data ...". The number of graphical displays per page is roughly one. The methodology is closely tied to the functional nature of the system measured. Derivatives are used where these have intrinsic meaning in the context.
FDA is however a very different type of book. This book is grounded in a classical, linear, tradition, a choice facilitating computation and sub- sequent analysis. Linear smoothers and multivariate techniques such as principal components are composed in clever and useful ways.

Contents:
1. Introduction
PART I : Introduction: Theory and Complexity
2. Duality theory for linear optimization
3. A polynomial algorithm for the skew-symmetric model
4. Solving the canonical problem
PART II : The Logarithmic Barrier Approach
5. Preliminaries
6. The dual logarithmic barrier method
7. The primal-dual logarithmic barrier method
8. Initialization
PART III: The Target-following Approach
9. Preliminaries
10. The primal-dual Newton method
11. Applications
12. The dual Newton method
13. The primal Newton method
14. Applications to the method of centres
PART IV : Miscellaneous Topics
15. Karmarkar's projective method
16. More properties of the central path
17. Partial updating
18. Higher-order methods
19. Parametric and sensitivity analysis
20. Implementing interior point methods
APPENDIX A: Some Results from Analysis
APPENDIX B: Pseudo-inverse of a Matrix
APPENDIX C: Some Technical Lemmas

The major areas of research in linear programming during the last decade have been interior point methods. This text is concerned with theory and development of interior point methods that have been shown to be polynomial. Part I is a self-contained treatment of linear optimization (used by the authors in preference to programming to avoid confusion with computer programming), including duality theory and a straightforward polynomial method: the Dikin Step Algorithm. A reader whose view of duality is intimately bound to the simplex method will find this presentation revealing. Part II is independent of Part I and it discusses the efficient algorithms, logarithmic barrier methods, that are used in commercial optimization soft-ware. Logarithmic barrier methods, also known as the central-path-following methods, are generalized as target-following-methods in Part II. Chapter 20 tackles some of the implementation issues including preprocessing, solving sparse linear equations, parameters, and identifying an optimal basis. There is a list of available software associated Internet addresses.
The great strength of this work is its unifying treatment of what has seemed to be a widely disparate and escalating field. The authors bring to the fore the role of duality and strictly complementary solutions in the development of the algorithms. Mathematical programmers trained in the classical theory of linear optimization culminating in the simplex algorithm and its many variants may find interior point methods dauntingly difficult. You should be re-assured by the authors that if you have mastered the algebras of pivoting, then the tools needed for interior point methods are only slightly more advanced. If you are willing to make the effort, I highly recommend this book for its well structured, strongly motivated, and clear presentation of a field considered to many to be the most important advance in optimization since 1947.

Contents:
SECTION 1: Forerunners
Abbe, Achenwall, Arbuthnot, Bayes, Bernoullis, Bienaymé, Boscovich, De Moivre, De Monmort, Déparcieux, de Witt, Graunt and Petty, Helmert, Huygens, Lambert, Laplace, Newton, Pascal, Quetelet, 'sGravesande, Sinclair, Süssmilch, Wargentin
SECTION 2: Statistical Inference
Birnbaum, Cox G., Cramer, David, de Finetti, Elfving, Fisher R.A., Galton, Guttman, Hoeffding, Hotelling, Hsu, Jeffreys, Kendall M.G., Kitagawa, Neyman, Pearson E., Pearson K., Pitman, Savage, Sverdrup, Thiele, Wald, Yule
SECTION 3: Statistical Theory
Aitken, Anderson, Bol'shev, Bortkiewicz, Bose, Chuprow, Cochran, Edgeworth, Gumbel, Hájek, Hartley, Langevin, Lévy, Lüroth, Rényi, Smirnov, Wilks, Wold, Wolfowitz
SECTION 4: Probability Theory
Bernstein, Bonferroni, Cantelli, Cauchy, Chebyshev, Feller, Gnedenko, Hannan, Jordan, Kolmogorov, Liapunov, Linnik, Markov, Ramsey, Wiener, Yanson, Yastremski
SECTION 5: Government and Economic Statistics
Bowley, Engel, Far, Fisher I., Franscini, Gini, Hansen, Kiaer, Konüs, Körösy, Lexis, Mahalanobis, Nightingale, Slutskii, Westergaard, Willcox, Winkler
SECTION 6: Applications in Science and Medicine
Gosset, Goulden, Hill, Nemichinov, Snedecor, Spearman, Wilson, Yates
SECTION 7: Applications in Science and Engineering
Boltzmann, Deming, Gauss, Maxwell, Scott, Wilcoxon
APPENDIX: A Statistical Analysis of Life Lengths of Leading Statistical Personalities
"It may happen that a certain person may also be assigned to some other section, ... The numbers of these additional sections are indicated. The name of the author of an entry is (usually) indicated at the end. Unsigned entries (many of the longer ones) have been composed by the editors..."

Readership: Anyone interested in statistics. Those unfamiliar with particular topics may need to refer to the editors' Encyclopedia from which most of these biographies are extracted.

Those who have the Encyclopedia to hand will nonetheless find the format and the additional entries make this volume well worth possessing.

Contents:
1. The search for causes: An overview
2. The controlled randomized experiment
3. Inferring causes from observational studies
4. Epidemiological studies
5. Summing up: Replication and triangulation
6. Coincidence and significance
7. Sampling
8. Content analysis
9. Surveys and change of venue
10. Trademark surveys: Genericness
11. Trademark surveys: Confusion
12. The jury: Composition and selection
13. DNA profiling: Probabilities and proof

Readership: Judges, lawyers, academics, law students, legislators, the business community and the public at large

The objective of this book is to explain statistical concepts which are useful in legal situations to those without a formal statistical background. It is devoid of mathematics (although not numbers), which is entirely appropriate for its intended audience. It includes over twenty case histories, spanning a wide range of areas and issues, and including analyses of the death penalty, jury selection, mass torts and vac-cine evaluations. Such a book is very timely. There has been a recent growth of recognition of the relevance and applicability of statistical arguments in legal fields-and also of the unfortunate ignorance about this amongst those in the legal profession. The book deserves, even needs, to be brought to the attention of law students and the judicidiary.
Apart from the evident value this book will have for its intended audience, it makes stimulating light reading for the statistician keen to see statistical ideas applied in relatively novel situations. Indeed, the appearance of this book makes one wonder if legal applications of statistics might someday have an impact on the development of future statistical methodology, in the same way that agricultural and medical applications have in the past.

Contents:
1. Mad cows or crazy communications?
2. A diagnostic for risk coummnications failures
3. Dioxins, or chemical stigmata
4. Hamburger hell
5. Silicone breasts
6. Lost in regulatory space: rBST
7. Gene escape
8. Mother's milk
9. Ten lessons

Readership: Scientists interested in public acceptability of science, journalists

As a term of art, "risk communications" has been around since the mid-1980s. Yet participants at a risk conference last year at Queen's University made clear that too few in government and the sciences take risk communications seriously. Mad Cows and Mother's Milk should help rectify that problem. Douglas Powell and William Leiss present scrupulously fair case studies of seven failures in risk communications C mad cow disease, dioxins, a particular strain of E. Coli, silicone breast implants, recombinant bovine growth hormone, plant biotechnology and PCBs.
Their conclusion: In the United States and Britain, the problem has largely been poor risk communications by public officials; in Canada, it is largely the lack of any risk communications. The authors summarize their advice in ten lessons for regulators which really break down to three: earn and retain the public's trust by willingly publicizing all relevant information, remember that there is always more to risk than the strictly scientific and never, ever claim there is no risk. Government and the beef industry, of course, trashed all these lessons in the case of mad cow disease and Britain is still paying the price. (The authors contend that the first general air-ing of mad cow disease in Canada occurred in a December 4, 1995 article in the Vancouver Sun; however, earlier mentions occurred in the Calgary Herald and Montreal Gazette. As well, the Canadian Press devoted a brief to the subject on May 12, 1995. If Internet discussion groups count in risk communication, so should a major news wire.)
While Powell and Leiss convincingly demonstrate the essential role of risk communications as a "causeway" for all aspects of risk management, they touch only lightly on how to improve the mass media transmission of such communications. Since they rightly note that many supposed "media analysts" have little practical experience C and demonstrate several times their own limited understanding of the realities of daily news operations C perhaps such omission is the wisest course. It leaves the door open for a follow-up work.
More disturbing is another self-imposed limitation of this volume C the decision not to ascribe blame. Recent public inquiries in Canada and France have shown that public officials entrusted with the safety of a national blood supply are capable of knowingly imperiling public health to save money and then destroying documents to conceal their misdeeds. This is not simply poor risk communications; this is immorality on a grand scale. And so were some of the case studies in Mad Cows and Mother's Milk, nothing is to be gained by ignoring the immorality of those scientists who knew and kept silent.

The editors of this encyclopedia which consists of almost five thousand pages should be commended for their fine work. These generals managed to marshall an army of more than eight hundred contributors in such an efficient way that the work was started and published within three years. The six volumes should be in every library.

Contents:
SECTION I : Graphical Failures
1. How to display data badly
2. Graphical mysteries
SECTION II : Graphical Triumphs
3. Graphical answers to scientific questions
4. Three graphic memorials
5. A Nobel graph
6. Todai Moto Kurashi
7. Picturing an L.A. Bus Schedule
SECTION III: Graphical Forms
8. Humble pie
9. Double Y-axis graphs
10. Tabular presentation
11. A rose by another name
12. Trilinear plots
13. Implicit graphs
SECTION IV : Using Graphical Methods
14. Measuring graphicacy
15. Graphs in the presidential campaign: Why weren't
they used more broadly?
16. Visual aids when comparing an apple to the stars
SECTION V : Improving Graphical Presentations
17. Integrating figures and texts
18. Elegance, grace, impact, and graphical displays
19. Sense-lining
20. Making readable overhead displays

Readership: The general public, journalists, educators, producers and users of data

Does the world need another general book on statistical graphics? With Tufte's sequence of books celebrating the wonder of good quantitative displays, one might have thought not. Yet, this book seems not only necessary but long overdue. Written with humour in a down-to-earth style, Wainer's book is the best introduction to statistical graphics for the general public. The author reworks a collection of articles which he wrote for Chance magazine and knits them together to produce a lively and entertaining introduction to the subject. All the classics are there (for example, Napoleon's Moscow campaign, London's cholera outbreak), as well as some important modern stories (for example, the Challenger disaster) and a few new ones (for example, the Mann-Gulch fire). Dip in any-where and a fascinating graphical story unfolds. Most important for a general audience, the stories are short and to the point. Readers will no doubt have favourites to use in class or to recommend for the benefit of others.
Blemishes, for a book on visual display, include colour plates separated far from the related text (presumably beyond the author's control), arbitrary changes in font size within numerical tables, and at least one transcription error between two tables. But these are minor compared to the book's important potential to reach a wide audience outside the statistical profession. I highly recommend this book. It is the book on graphics for everyman.

Table des matières:
1. Accroissement et inertie des populations
2. L'analyse des événements
3. Méthodes statistiques
4. A la recherche des causes
5. Les évolutions des populations

Dans ce livre les auteurs nous offrent une initiation à la démographie, une discipline qui depuis le début (avec la table de John Graunt en 1662) est très liée avec la statistique. La démographie étudie l'évolution quantitative d'une population et les éléments qui sont à la base de cette dynamique. Les outils traditionnels (table de mortalité, pyramide des âges, ...) et les méthodes statistiques sont présentés sans trop de mathématiques. Il y a beaucoup de tableaux et graphiques, surtout sur la situation démographique en France. En conclusion, ce livre nous donne une excel-lente introduction à une discipline extrêmement intéressante.

Contents:
1. Basic principles of population genetics
2. Counting methods and the EM algorithm
3. Newton's method and scoring
4. Hypothesis testing and categorical data
5. Gene identity coefficients
6. Applications of identity coefficients
7. Computation of Mendelian likelihoods
8. The polygenic model
9. Markov chain Monte Carlo methods
10. Reconstruction of evolutionary trees
11. Radiation hybrid mapping
12. Models of recombination
13. Poisson approximation
APPENDIX: Molecular Genetics in Brief

Modern genetics is an exciting and rapidly developing area of science. There is currently an explosion of molecular data, coupled with a continual turnover of experimental techniques and hence types of data and questions of interest. There is a long tradition of mathematical and statistical involvement in the subject, and it remains a fertile, if under-populated, area for statisticians. The breadth of the area, and its pace of change, make a comprehensive treatment of statistics in modern genetics impossible. This book aims to give both the flavour and details of many of the applications, and in this it is successful. It also exhibits the considerable diversity of statistical and mathematical methods which prove helpful. To borrow a metaphor from the geneticists, the field of genetics is a good model system for learning modern computationally intensive statistical methods.
Aside from being a valuable reference for those already in the field, the book (complete with numerous exercises) will be a good teaching tool, and an accessible introduction for newcomers to the area.

Contents:
1. Basic concepts in survival analysis
2. Estimation of functions and parameters
3. Comparison of survival distributions
4. Correlation and regression analysis
APPENDIX A: Kidney Transplant Data
APPENDIX B: Haemodialysis Data
APPENDIX C: Ventilation Tube Data

This book offers an introduction to elementary methods for the analysis of failure-time data. Topics include the Kaplan-Meier estimator, the log-rank and other commonly used two-sample tests and Cox's regression model. There are numerous examples taken from the public health and medical fields. The material is presented at a basic (non-calculus) level, making it understandable to and suitable for students and re-searchers in health-related fields such as epidemiology, environmental health sciences and the medical sciences who want a "how to" introduction to this important area. Because of the limited technical detail and discussion of the properties of methods, the text is less well suited to students in biostatistics and statistics who wish to have a more rigorous introduction to failure time methods.

Contents:
1. Introduction
2. Statistical concepts
3. The design of clinical trials
4. Multi-arm trials
5. Interim analysis and data monitoring committees
6. Data management and quality control
7. Reporting of results
8. Pitfalls
9. Exploratory analyses
10. Summary and conclusions

The authors set as the aim of this book to improve the mutual understanding by clinicians and statisticians of the principles of cancer clinical trials. This aim is met very well as the book provides a comprehensive account of all aspects of cancer clinical trials written in a manner accessible to non-statisticians. The book describes statistical concepts of clinical trials and gives practical advice in how to design and conduct a successful trial, and how to avoid the most common pitfalls. Discussed issues are illustrated in detail by examples from trials run by the Southwest Oncology Group. The fact that all royalties from the sale of this book will go to the Southwest Oncology Group Statistical Center may be yet another reason to buy this excellent compendium.

Contents:
1. Introduction
2. Study design and planning
3. Data definition, forms and database design
4. Computers in clinical trials: Hardware, operating systems, and database management systems
5. Data entry and distributed computing
6. Patient registration
7. Local data and management systems
8. Central quality control of data
9. Data management and good clinical practice
10. Software tools for trials management
11. Follow-up and close-out phase
12. Training and education

This book thoroughly examines the practicalities of data management within clinical trials. The book covers the entire duration of a trial from protocol development to final analysis and addresses the steps to take, and pitfalls to avoid in the collation, recording and processing of data. Specific topics include the design and production of paper records, administrative systems, computer hardware and software, data entry, verification, quality control and good clinical practice. The role of the data co-ordinating centre is also reviewed, together with an outline of the roles and potential training requirements of key staff. Throughout there is an emphasis upon the value of thorough planning and consistent procedures, and the themes covered would be relevant to a wide variety of clinical trials regardless of size or subject. The book will provide a useful reference to anyone involved in trial management, and be of particular value to those new to the field.

Contents:
1. Introduction
2. Basic considerations in setting standards
3. The pollutant-effect relationship and other links
4. Current and developing incorporation of uncertainty and variability in standard-setting
5. Current standards: Examples
6. Conclusions: The current situation and a forward look

Readership: Statisticians and environmental agency personnel involved with standard setting

This is the final report of a critical review of the importance and current state of treatment of uncertainty and variation in environmental standard setting, prepared for and at the request of the Royal Commission on Environmental Pollution. Standard setting in general is considered, with the term standard being given a broad definition and specific examples used for illustration. Terminology is defined and the representation of uncertainty and variation in the context of aims and objectives, pollutant-effect relationships, the formal standard, the criterion for monitoring or testing compliance and considerations of costs and benefits are discussed. It is concluded that, in general, existing standards do not adequately account for uncertainty and variation. The statistically verifiable ideal standard with a non-prescriptive method of verification, is proposed as a preferred approach. It is emphasized throughout that statistics is the means of incorporating uncertainty and variation. The last chapter contains a call for action to make the need for the inclusion of statisticians known to those involved in standard setting and to raise the profile of environ-mental statistics in order to encourage the provision of an adequate supply of expertise.
For environmental statisticians, the book will provide reinforcement of many of their own thoughts and initiatives, and, for other statisticians, it will pro-vide an introduction to the topic in general, although it necessarily gives a limited and selective list of examples. Non-statisticians can easily choose parts of interest because of the many subsections and informative titles, should the more abstract discussion of types of standards and the reiteration of some points seem esoteric.

Contents:
1. Introduction to experimental design
2. Comparative experiments
3. Blocking
4. Factorial experiments
5. Screening experiments
6. Optimization methods

Readership: Working engineers and engineering students who need to understand statistical tools to improve manufacturing processes

Essentially this is a re-packing job with style and class. The author (who works for Intel Corporation) has read several books on experimental design and has distilled certain topics useful for engineers, for example, one-way and two-way ANOVA, two-level factorial, fractional factorial and Plackett and Burman designs, blocking, steepest ascent and second-order surface fitting, into a small "nutshell" text. It was puzzling at first. Reading on p.56 that the "exact binomial test" could be used, I searched in vain for a reference to the binomial in the index and table of contents. Could this be in the "first book in this series", whose title is mentioned in the preface? After further search, I found an SAS program on pp. 300-301. I then realized that the SAS programs were to be used to perform the computations needed throughout the text. The book has several nice features, for example, many layouts of two-level designs, and triangular diagrams that show the aliasing for specific 2k-p designs. The writing is very clear. The problems (exercises) are cute. There is an air of friendly common sense throughout. This makes it an excellent introductory
"cook book" for statistically inexperienced engineers.

Contents:
PART I : Introduction to Shrinkage Estimates
1. Introduction
2. The Stein paradox
3. The ridge estimators of Hoerl and Kennard
PART II : Estimation for a Single Linear Model
4. James-Stein estimators for a single linear model
5. Ridge estimators from different points of view
6. Improving the James-Stein estimator: The positive parts
PART III: Other Linear Model Setups
7. The simultaneous estimation problem
8. The precision of individual estimators
9. The multivariate linear model
10. Other linear model setups
11. Summary and conclusion

Readership: Those interested in understanding, or extending their knowledge of, shrinkage estimation

The reference section of this well-written book has about five-hundred entries. The author has done an excellent job of summarizing and explaining all this material in six hundred and thirty-two pages. After some preliminaries, the author sets off on the historical survey of Section 1.4, which consists of forty-five pages with one-paragraph summaries of the main threads of the literature beginning with the Stein (1956) result that started it all. This very valuable survey sets the scene for the excursions into the main avenues of the chapter topics. The emphasis is mostly on theoretical developments; numerical illustrations appear throughout, but experimental consequences of calculations are not featured. Thus, for industrial users, this would be a reference volume, rather than a handbook. The text would be ideal for a specialized graduate seminar on shrinkage, and as a reference text for courses in regression. It is an essential library purchase.

Contents:
1. Density estimation for exploring data
2. Density estimation for inference
3. Nonparametric regression for exploring data
4. Inference with nonparametric regression
5. Checking parametric regression models
6. Comparing curves and surfaces
7. Time series data
8. An introduction to semi-parametric and additive models

This must be a very attractive book: when it was lying on my desk while preparing this review, it was constantly taken away by students and colleagues who were attracted by the topic and the nice presentation with graphics, examples, S-Plus material, etc. The book is in the first place a very practical introduction to nonparametric smoothing in the two important fields of density functions and regression curves. A glance at the more than two-hundred references reveals that most of them date from the nineties and hence it becomes clear that this is an up-to-date book with the most recent state of the art. Most technical details are left out, which makes the text accessible to non-mathematical readers. There is a rich choice of examples, exercises, hints for further reading and S-Plus illustrations. Compared to the several other recent books in this area, the present monograph has the advantage of being introductory and practical within a very reasonable number of pages.

Contents:
1. Introduction
2. The mathematics of probability sampling designs
3. Distributions induced by random sampling designs
4. Design-based estimation for general finite population quantities
5. Inference for descriptive parameters
6. Analytic uses of survey data
7. Sampling strategies in time and space

This is a clearly and carefully presented development of the theory of sampling. Although the ideas contained in this text are applicable to many areas of experimental science, it does not treat the practical issues of survey design or provide details of how to analyze data from samples. There are no il-lustrations using data and exercises are restricted to a few theoretical problems at the end of Chapter 2. It is, as the title states, a theoretical treatment which focuses on aspects of sampling theory not commonly found in other texts.
The short introductory chapter is followed, in Chapter 2, by a discussion of the mathematics of randomized sampling designs, the Horvitz-Thompson estimator and other linear estimations for population totals. Estimation in multi-stage sampling and unequal probability designs are also described. In Chapter 3, the limiting distributions of standardized and studentized estimators of totals and means are discussed, using the finite population central limit theorem and formal Edgeworth expansions in the case of simple random sampling. Saddle point approximations to the distribution of the sample sum are explored, and both saddle point and bootstrap resampling methods are used for confidence interval construction.
Chapter 4 extends the development of design-based confidence intervals to the estimation of parameters for simple and complex designs. The treatment of means, ratios and quantiles is unified through an estimating function formulation. Different structures for these estimating functions lead to a wide range of theoretical results. Descriptive sampling inference is covered in the next chapter, in which the idea of superpopulation models is introduced, and Chapter 6 discusses the meaning of analytic inference. The final chapter considers survey populations where the units are labelled by times or locations in space. More than two hundred references are cited.
This text could be used for a graduate course on sampling theory, but would need to be supplemented by other texts covering practical illustrations and examples.

Contents:
1. Bayesian foundations
2. A non-informative Bayesian approach
3. Extensions of the Polya posterior
4. Empirical Bayes estimation
5. Hierarchical Bayes estimation

Readership: Researchers and practitioners in sample surveys and finite population sampling

This monograph provides a detailed study of Bayesian inference for finite population sampling. Although the authors admit that their aim was not to provide a comprehensive account of this area, they succeed admirably in providing a wealth of material which is not available in standard texts on Bayesian inference or sample surveys. Connections with frequentist design-based or model-based procedures are high-lighted, particularly in Chapter 2 where a noninformative approach based on the Polya posterior distribution is considered. This approach is then extended to allow various types of prior information to be incorporated. The final chapters concern small area estimation using hierarchical and empirical Bayes methods. Here, the level of algebriac detail presented is quite high. Indeed, the book is written at quite a high level throughout, and to gain full benefit from it, the reader requires some prior knowledge, both of Bayesian inference and of sample survey methods.

Contents:
1. Data-driven inference
2. The bootstrap
3. Outliers contamination and robustness
4. Location tests for two independent samples
5. Location tests for single and paired samples
6. More one- and two-sample tests
7. Three or more independent samples
8. Designed experiments
9. Correlation and concordance
10. Bivariate regression
11. Other regression models and diagnostics
12. Categorical data analysis
13. Further categorical data analysis
14. Data-driven or model-driven?

Readership: Students with a preliminary grounding in the subject, research workers, scientists, engineers, industrialists, technologists, or managers in any field who have a basic knowledge of statistics

The term 'data-driven' invites comparison with the term 'evidence-based' currently popular in medicine C and likewise prompts the question about what the alternative might be. Here it is intended to distinguish the methods described in the book from 'model- driven' methods. The latter are to distinguish the methods described as being based on using probabilistic or mathematical models to encapsulate the main features of the data, while data-driven methods are based on less specific and restrictive models (for example, merely that the data have been randomly drawn from a symmetric distribution, or that two samples have been randomly drawn from populations which are identical apart from their locations, and which are otherwise unspecified). There is no distinct boundary between the two approaches, but rather a difference in emphasis. Data-driven methods are a more modern development, and one for which computer power has been a key motivator. This is illustrated by the author's list of the main themes covered in the book: randomization and permutation tests, bootstrap methods, robust estimation and treatment of outliers, diagnostics, and inter-relation-ships between inference methods. The book is a nice introduction to modern methods in applied statistics.

Contents:
1. Introduction
2. The basic bootstrap
3. Further ideas
4. Tests
5. Confidence intervals
6. Linear regression
7. Further topics in regression
8. Complex dependence
9. Improved calculation
10. Semi-parametric likelihood inference
11. Computer implementation

This book is intended to give a balanced account of resampling methods, to provide the theoretical underpinnings of the methods and to illustrate the techniques with a large number of applications. The coverage is comprehensive, making the book very useful for anyone interested in using resampling techniques in a variety of settings. The index of examples has approximately two-hundred entries. There is an extensive set of problems at the end of each chapter. This is followed by a set of practical examples based on the S-plus functions and sets of data included with the book and available from the home page for the book, http://dmawww.eplf.ch/davison.mosaic/BMA/. The book is well written and is at a level which ensures its usefulness for a wide range of readers.

Contents:
1. Stochastic simulation
2. Bayesian inference
3. Approximate methods of inference
4. Markov chains
5. Gibbs sampling
6. Metropolis-Hastings algorithms
7. Further topics in MCMC

The sub-title is perhaps more appropriate than the main title, since Markov Chain Monte Carlo (MCMC) methods per se occupy less than half of the text. In any case, I expect that the book will, deservedly, sell very well. MCMC methods are currently revolutionizing many branches of applied statistics, and their proper-ties are occupying the attention of many researchers in statistics. Although there already exist excellent monographs conveying the fruits of this research to a wider audience, there is an ongoing need for more of them, filling different niches, and making accessible the latest research developments. The present text is aimed at a more introductory level, and is embedded in a wider range of related theory, and the Bayesian paradigm for statistical inference, making it appropriate for a postgraduate course text. Recent research results are presented, although usually briefly and in general terms. An extensive list of references points to the literature, and an author index is provided in addition to a subject index.
There are three chapters on MCMC algorithms. Gibbs sampling and Metropolis-Hastings algorithms are given roughly equal emphasis with a chapter each. This reflects a compromise between the early prominence of the former, and the wider availability of theoretical results for it, whereas the more general Metropolis-Hastings algorithms are becoming more prominent in practice, as more challenging applications are being tackled. A final chapter on advanced topics primarily covers model choice, with a brief discussion of convergence acceleration via variable-direction transit-ions and auxiliary variables.
The style is accessible and production standards are high: I noticed only one technical error and very few typographical errors. Given its modest price, it is a must for every research library, and should be given serious consideration for use as a graduate text.

Contents:
1. Proximity data
2. Measures of similarity, dissimilarity and distance
3. Spatial representation of proximity data: Metric and non-metric multidimensional scaling
4. Interpreting, diagnosing and comparing multidimensional scaling
5. Three-way multidimensional scaling
6. Asymmetric and rectangular data
7. Tree models for proximity data
APPENDIX A: Distances in classical multivariate analysis
APPENDIX B: Software for multidimensional scaling

Readership: Experimental scientists, statisticians, students of multivariate analysis

This book is Volume 4 in Kendall's Library of Statitics. It describes the many alternative ways for constructing measures of similarity and dissimilarity, and then presents a range of multidimensional scaling techniques and tree models for illustrating these measures. (This reviewer found useful Krzanowski's (1988, p.164) use of Procrustes analysis to compare configurations resulting from different dissimilarity measures.) Much of the material is available in general text books on multivariate analysis, but the authors have usefully brought the subject up-to-date; some thirty-two of the references have been published since 1990; see, for example, Zielman and Heiser (1996). Based on Carter (1996), the material on tree models is refreshing and interesting. It is suggested that hierarchical clustering algorithms may often provide good approximations. Techniques are beautifully illustrated on a large number of sets of data, which have been expertly chosen from a wide range of areas. In the extensive software Appendix reference is made to an S-Plus function for non-metric scaling, but there is no mention of how easy it is to program metric scaling in S-Plus.
Carter, J.E. (1996). Tree Models of Similarity and Association. Beverly Hills, California: Sage.
Krzanowski, W.J. (1988). Principles of Multivariate Analysis. Oxford University Press. [Short Book Reviews, Vol. 8, p.41]
Zielman, B. and Heiser, W.J. (1996). Models for Asymmetric Proximities. British Journal of Mathematical and Statistical Psychology 49, 127-146.

Contents:
1. Introduction to spatial data and S+SpatialStats
2. Getting started with S+SpatialStats
3. Visualizing spatial data
4. Analyzing geostatistical data
5. Analyzing lattice data
6. Analyzing spatial point patterns
7. S+SpatialStats and GIS
APPENDIX A: Installing S+SpatialStats
APPENDIX B: S+SpatialStats Index by Category
APPENDIX C: Sample Data Sets
APPENDIX D: Function Reference

Readership: Statisticians, scientists and other data analysts who have a working knowledge of S-Plus and some familiarity with spatial statistics

S+SpatialStats is an add-on package (sold separately) for the S-Plus system for data analysis and graphics. It provides a collection of functions for two-dimensional spatial data. The collection appears to be fairly comprehensive, as it provides methods for all three types of spatial data: geostatistical data, lattice data and spatial point patterns. Recent interest in spatial statistics and the availability of GIS databases will likely make this a popular addition.
The book is well laid out, similar in style to a number of LaTeX manuals, which I find aesthetically pleasing and easy to navigate. A concise introduction gives a good overview of the three types of spatial data and some associated spatial statistical concepts. Each chapter gives a brief survey of a few spatial data analysis topics and illustrates through a series of examples how to accomplish the analyses with S+SpatialStats and S-Plus functions. Examples are well illustrated with the graphics capabilities of S-Plus. The appendices provide the usual function lists and function documentation that a manual should have. If you like to use S-Plus and have spatial data, I highly recommend this book.

Contents:
1. Statistique d'ordre et de rang
2. Statistique descriptive: Une population, une variable
3. Simulation des données
4. Echantillonnage de populations finies et infinies
5. Données binaires: Le modèle binomial
6. Données binaires: Le modèle hypergéométrique
7. Approximations normales
8. Deux échantillons, données binaires: Le test exact de Fisher et autres
9. Données catégorielles: Tableaux de contingence
10. Données continues: Modèles non paramétriques
11. Régression multilinéaire

Ce livre élémentaire de la statistique concentre sur les modèles statistiques précis (par exemple, le binôme, le hypergéométrique et le normal) et leur importance pédagogique et pratique. Il ressort de la discussion au sujet de la simulation des échantillons prélevés au hasard au commencement du livre que l'accent d'intensité est sur les applications. La présentation est partisan des tests exacts plutôt que des méthodes d'approximation habituelles.

Contents:
1. Introduction to dispersion models
2. Natural exponential families
3. Exponential dispersion models
4. Tweedie models
5. Proper dispersion models

Readership: Lecturers, students and researchers concerned with generalized linear models, statisticians dealing with non-normal data

The author uses his own research insights to provide a very clear introduction to the theory of multivariate dispersion models and their use as error distributions for generalized linear models. He plans to deal with statistical analyses based on these models in a subsequent volume. The various classes of dispersion models in the book include distributions suitable for continuous, discrete and mixed data, on the real line and on the half line with and without zeroes, and also proportions and directional data. The first three chapters contain core material for the study of generalized linear models.
Clarification in Chapter 2 of the importance of the deviance and the variance function follows a general overview in Chapter 1. In Chapter 3, exponential dispersion models provide examples of both re-productive and addititve models; quadratic variance functions, saddlepoint approximations, and tail area approximations are also examined. The author moves on in Chapter 4 to consider exponential dispersion models that are closed under scale transformation or under translations (Tweedie models). Finally he gives a general definition of proper dispersion models, discusses methods for their construction, and looks at Studentization.
Throughout there are examples, and each chapter ends with notes and exercises (no solutions are given). There is plenty of material here for a graduate course in distributions for generalized linear models. I recommend supplementation with practical analyses of data and look forward to the promised companion volume.

Contents:
PART I : Modeling and Data Analysis
PART II : Statistical Inference in Parametric Models
PART III: Elements in Multivariate Analysis
PART IV : Topics in Insurance, Finance and Hydrology
PART V : Case Studies in Extreme Value Analysis
APPENDIX: An Introduction to XTREMES

There are quite a few books on extreme-value theory available now. The aim of the present book is not so much to provide a broad introduction to the field, but to focus on numerical methods. Many statistical techniques are discussed, most of them connected with extremal values, but many more are not. A separate chapter provides connections with the fields of insurance, finance and hydrology. The chapter on multivariate extremes is sketchy.

Contents:
1. Introduction
2. The relationship between finance and actuarial science
3. Actuarial applications and generalized linear models
4. Consumer credit and statistics
5. Methodologies for classifying applicants for credit
6. Credit scoring and quality management
7. Consumer credit and business cycles
8. Probability in finance: An introduction
9. Introduction to financial economics
10. American options
11. Notes on the term structure models
12. Default risk
13. Non-parametric methods and option pricing
14. Stochastic volatility
15. Market time and asset price movements: Theory and estimation

In the Introduction the editors write: "The book is a collection of chapters illustrating some of the problems and issues that arise in financial statistics. In particular, it is not a text book. It is in-tended to be a taster, showing the range and scope of financial statistics. In pursuance of this aim of illustrating the breadth of field, some of the chapters are technically quite demanding, while others are less so. This serves to demonstrate the scope for different kinds of statistical applications."
My impression of this collection of loosely related articles by nineteen authors is that there is only little about statistics in this book. It contains various papers on finance, economics, insurance, econometrics. Some of the papers give short introductions to these fields; some of them do not contain formulae, some of them require stochastic calculus on semi-martingale level. Most chapters are about probabilistic modelling, not about statistics.
Statistical problems in modern finance and insurance are related to the estimation and forecasting of correlations and variances, risk management (value at risk (VaR)), extreme values, estimation of parameters and coefficients of stochastic differential equations (SDEs). I did not find too much about these topics, and I also miss the relevant references, for example about the estimation of volatility. The simplistic maximum likelihood estimator of the volatility based on geometric Brownian motion observed at equi-distant points (see p. 285) is not what statisticians in banks would use. The entire work of the Aarhus school (M. Sorensen and co-workers) on estimation of coefficients for SDEs is omitted. There is discussion going on (between and among practitioners and theoretically oriented people) what VaR is and how should this notion be defined in order to make financial risks safer. I did not find much about this topic in the book, not even a clear definition or a discussion of various possible definitions of VaR. Closely related to VaR, but also to the problems of reinsurance (catastrophe insurance) is extreme value statistics. This topic is omitted as well.
Finally, this book contains some chapters which can be useful as an introduction to various topics of finance and insurance, but it is not a useful reference book on topics related to statistical problems.

Contents:
0. The model
PART I : Complete Markets
1. Pricing
2. Optimization
3. Equilibrium
PART II: Incomplete Markets
4. Hedging
5. Optimization
6. Pricing
7. Transaction costs
APPENDIX: Historical Notes

This text is to be situated between the very successful Brownian Motion and Stochastic Calculus, 2nd edition, 1991, by I. Karatzas and S.E. Shreve [Short Book Reviews, Vol. 8, p.45] and Methods of Mathematical Finance, to appear, by the same authors. Aimed at a more mathematically oriented audience, the book goes well beyond the usual frictionless market model by introducing consumption, transaction costs and portfolio constraints. Both complete as well as incomplete markets are considered. The author mainly follows the theorem-proof style, with many exercises dispersed throughout. The young researcher/postgraduate student will be able to glance at the forefront of current research in mathematical finance. The author's clear and careful writing makes reading a pleasure. A lot of material, hitherto available only in research papers, will now reach a wider audience. This is a most useful addition to the fast growing literature on mathematical finance.

Contents:
PART I : Background
1. Introduction and overview
2. Foundations of financial economics
PART II : Methodological Foundations
3. Variational inequalities
4. Projected dynamical systems
5. Nonlinear networks
PART III: Single Country Models
6. Static single country models
7. Static single country hedging models
8. Dynamic single country models
9. Static imperfect market models
10. Dynamic imperfect market models
PART IV : International Models
11. International financial models
12. International models with hedging
13. Imperfect market models
PART V : Flow of Funds and Estimation
14. Flow of funds models
PART VI : Empirical Results
15. Empirical analysis

Readership: Finance students, practitioners, theorists, management scientists, computational scientists, control theorists
Traditional finance is concerned principally with the temporal allocation of resources, but the existence of distinct economic sectors and the multiplicity of financial instruments call for studying capital flows between markets as well as over time. Early programming approaches (such as those pioneered by A. Charnes and W.W. Cooper) focused mainly on the temporal allocation problem. The present book examines both temporal and geographical allocation problems, viewing multisectoral, multi-instrument financial al-locations as network flows. Finite-dimensional variational inequality theory is used to address the existence and uniqueness of static equilibria, while dynamic systems theory is used to study disequilibrium and convergence behaviour. Empirical applications are illustrated using such contexts as estimating funds flows.

Contents:
1. Probability
2. Random variables and expectation
3. Distributions, transformations and moments
4. Convergence concepts
5. Statistical inference

Any (mathematical) statistician or probabilist, student or researcher alike, who browses through this text will give a similar judgement: a fairly standard introduction to probability and statistics using measure theoretic language. The fact that the author stresses "econometric theory" in the title however reflects the very specific intended readership (students in economics). Modern econometrics increasingly relies on more advanced stochastic models. This leads to a need for introductory texts in stochastics which yield the envisaged students a sounder basis. In that respect, the book is excellent. All concepts introduced are clearly defined, presented in a mathematically correct way and many exercises are provided. Too often in applied fields mathematical correctness is sacrificed for so-called applied (intuitive) under-standing. Though the latter is important, it is my firm belief that without the former, the solutions to real applied problems in the field of econometrics become shaky. The author is, therefore, to be congratulated to have taken up this view and through this book will, I hope, increase among economics students the deeper understanding of stochastics. I for one would very much like economics students interested in finance to have this book under their (scientific) belt.

Contents:
1. Introduction to probability theory
2. Arrivals and services
3. Queueing systems: Average performance
4. Queueing networks
5. Markov chains and processes
6. Queues in Markovian environments

This book has a narrower scope than its title would suggest, concerning itself mainly with queueing models arising in engineering, with no modelling of financial data, insurance risk, genetics or biological processes. An effort has been made to make the material easily accessible, though the presentation is very restrained and understated, and could have been made more lively. The informality might make the book attractive to non-mathematicians, but I can see even these readers bemoaning the absence of simulation modelling and testing. As it is, results presented as theorems lack a clear statement of conditions and have proofs so informal that they should not have been honoured with a Q.E.D.

Contents:
1. Introduction
2. A brief and superficial history of statistics for drug developers
3. Design and interpretation of clinical trials as seen by a statistician
4. Probability, Bayes, p-values, tests of hypotheses and confidence intervals
5. The work of the pharmaceutical statistician
6. Allocating treatments to participants in clinical trials
7. Baselines and covariate information
8. The measurement of treatment effects
9. Demographic subgroups: Representation and analysis
10. Multiplicity
11. Intention to treat
12. One-sided and two-sided tests
13. Determining the sample size
14. Multicentre trials
15. Active control equivalence studies
16. Meta-analysis
17. Cross-over trials
18. N-of-1 trials
19. Sequential trials
20. Dose-finding
21. Concerning pharmacokinetics and pharmacodynamics
22. Bioequivalence studies
23. Safety data, drug monitoring and pharmaco- epidemiology
24. Pharmaco-economics and portfolio management
25. Glossary

This is a highly readable book, written by a statistician who has worked in the pharmaceutical industry, and has the rare ability to communicate comprehensibly to non-statisticians. Although aimed at the non-statisticians in the pharmaceutical industry, I think it will also be of interest for statisticians both within and outside this specific arena. The messages from the wide range of topics covered are reinforced by the use of boxed keypoints, interesting chapter openings, as well as humour. The glossary is useful, and there is a comprehensive list of references after each chapter. I recommend the book highly.

Contents:
1. Basic probability and statistical distributions
2. Fundamentals of statistical inference
3. Fundamental issues in experiment design
4. Data analysis of treatment-versus-control differences
5. Treatment-versus-control multiple comparisons
6. Trend testing
7. Dose-response modelling and analysis
8. Introduction to generalized linear models
9. Analysis of cross-classified tabular/categorical data
10. Incorporating historical control information
11. Survival-data analysis

Readership: Postgraduate students and researchers in environmental biology or toxicology; teachers of statistics

From the chapter and sections headings alone, it would be difficult to identify this book as anything other than an introductory textbook in statistics with a slightly unusual but modern-sounding selection of topics. I feared therefore that this was to be yet another example of the Statistics for Topic X genre in which a standard introductory statistics book has had some examples from Topic X grafted onto the end of each chapter, so capturing a `new' market with little extra effort. It was therefore a pleasure to find that the biology and toxicology was thoroughly integrated with the chosen statistical techniques throughout most of this book. Students in this subject area would definitely need the support and encouragement offered by this integration with their own subject, since the book tackles a variety of mathematically sophisticated statistical methods in a condensed but admirably clear and sound manner. SAS code and output is utilized extensively in the text. Teachers of statistics to students from other disciplines could well find this book useful, both for its condensed summaries of some of the more sophisticated techniques in particular that for generalized linear models is helpful hand also for the copious exercises and worked examples.

Contents:
PART I : Random Series and Linear Transformations of Sequences of Independent Random Variables
0. Random elements and their convergence (preliminary notions)
1. Series of independent random elements
2. Linear transformations of independent random elements and series in sequence spaces
PART II : Limit Theorems for Operator-Normed Sums of Independent Random Vectors and Their Applications
3. Operator-normed sums of independent random vectors
4. Operator-normed sums of independent identically distributed random vectors
5. Asymptotic properties of Gaussian Markov sequences
6. Continuity of sample paths of Gaussian Markov processes
7. Asymptotic properties of recurrent random sequences
8. The interplay between strong and weak limit theorems for sums of independent random variables

This scholarly written book gives a comprehensive coverage of the asymptotic properties of sample paths of linearly transformed sums of independent random variables, random vectors and random elements taking values in a topological vector space. The origins of the subject can be found in extensions of strong limit theorems to weighted sums in generalized summability schemes and in stochastic approximation. The book covers these topics with great respect for methods and completeness. Moreover, the authors push the subject much further by considering strong limit theorems for operatornormed random vectors; the study is based on a systematic use of general linear trans-formations in sequence space.
The book contains a wealth of new material apart from encompassing material from more than two-hundred scientific books and papers. In the comments the authors provide bibliographical notes to the individual chapters.

Contents:
1. The likelihood approach
2. Nonparametric methods
3. Linear models
4. Dependent observations

Readership: Research mathematical statisticians and theoretical biostatisticians

The subject of change-point analysis has been important in statistics for many years, but only recently have unified and intensive studies been attempted. Significant interest and activity have occurred in recent years, utilizing the full scope of theoretical tools. In this volume, the authors provide an impressive coverage of limit theorems and approximation techniques available for the study of likelihood ratio methods within change-point models. These include single and multiple changes in means through non-parametric estimators, trend changes in regression models and changes in means in ARMA processes. Though the book is primarily a technical presentation of relevant limit theorems, the authors have applied some of their results to a number of interesting real sets of data. Overall, this is a carefully prepared text, that provides a useful and timely overview of powerful asymptotic techniques that will be a valuable resource to researchers in the area.

Contents:
1. Artificial randomness
2. Stochastic models
3. Stochastic processes
4. Evaluation of statistical procedures

This book is a translation of a four-part 1995 German monograph on stochastic simulation, although the authors disagree with appropriateness of this terminology. It discusses a number of traditional topics such as the design and testing of uniform and non-uniform pseudo random number generators, simulating Markov chains, birth and death processes, Brownian motion and diffusion. However, in addition it presents some less common examples, including controlling traffic lights, kinetic gas theory and experimental assessment of various statistical results, including the Neyman-Pearson lemma, Bayesian decision theory and the Wald sequential test. The strength of the book lies in the computer experiments which can be conducted using the CD ROM and play a prominent part of the text. They pro-vide interesting laboratory exercises which could ac-company a simulation course at a senior undergraduate level. While not a complete reference on stochastic simulation, this book is highly recommended as a more experimental companion to one of the more comprehensive texts.

Contents:
PART I : Introduction
1. A chronicle of system design limitations
PART II : Limitations in Linear Control
2. Review of general concepts
3. SISO control
4. MIMO control
5. Extensions to periodic systems
6. Extensions to sampled-data systems
PART III: Limitations in Linear Filtering
7. General concepts
8. SISO filtering
9. MIMO filtering
10. Extensions to SISO prediction
11. Extensions to SISO smoothing
PART IV : Limitations in Nonlinear Control and Filtering
12. Nonlinear operations
13. Nonlinear control
14. Nonlinear filtering

Readership: Control theorists, practising control engineers. A fairly wide background in the control literature is presumed

There is a conventional list of desirable attributes of well-designed feed-back control systems, including such notions as fast rise time, low over- shoot, exemplary tracking performance, strong disturbance rejection, and admirable robustness properties.
The theme of the book can be taken as showing that the inherent structure of feedback systems imposes tradeoffs between these desirable design objectives.
The origin of the ideas goes back to Bode's work on telephone amplifier designs, but the text covers extensions of the ideas to multiple input-output systems, filtering problems, and some nonlinear system models.

Contents:
PART I : Introduction
PART II : Sampling and Monitoring of Pollution
PART III: Radiation
PART IV : Air Quality
PART V : Water Quality
PART VI : Health
PART VII: Policy Management

This is the third volume in the sequence titled 'Statistics for the Environment' [Short Book Reviews, Vol. 14, p.12; Vol 15, p.3]. It contains six-teen of the papers presented at the SPRUCE 3 environ-mental conference held at Merida in Mexico during 1995. As with the other two volumes, this is a well-edited package, with an integrated and readable set of papers presented in a consistent style. It should be of interest to both statisticians looking for sophisticated examples of statistical modelling, and to environmentalists looking for assistance from statistics.

Contents:
1. Regional frequency analysis
2. L-moments
3. Screening the data
4. Identification of homogeneous regions
5. Choice of a frequency distribution
6. Estimation of the frequency distribution
7. Performance of the regional L-moment algorithm
8. Other topics
9. Examples
APPENDIX: L-moments for Some Specific Distributions

Readership: Environmental scientists, hydrologists, civil engineers concerned with extreme value distributions; applied statisticians

Data relating to extreme environmental events are almost by definition scarce. Regional frequency analysis involves combining data from different sites for the estimation of the distribution of a positive random variable, with particular interest in its upper tail. The L-moments are shape parameters.
If x(u) is the quantile function (the inverse function of the distribution function) of a random variable, the L-moments are the integrals of x(u) weighted by the shifted Legendre polynomials in u (orthogonal on [0,1]). The L-moments can be expanded as linear combinations of order statistics. The ob-served order statistics give sample moments which form the basis of the estimation. Parametric models of the quantile function have been developed which show skew-ness and heavy tails. There may be no explicit expression for the probability density function. The five parameter Wakeby distribution is of this type. It exhibits a wide range of shapes, and includes the generalized Pareto distribution.
The statistical problems of combining data from various sites are treated carefully, addressing the preliminary screening of data, heterogeneity, etc.
Most of the references are to the hydrology literature. The authors might have drawn attention to techniques being used in the analysis of financial and insurance risk data. The potential use of regional frequency analysis to financial risk modelling is why I have extended the readership list to include applied statisticians.

Contents:
1. Plan of the book
PART I: Direct Probability, 1750 – 1805
2. Some results and tools in probability theory by Bernoulli, de Moivre, and Laplace
3. The distribution of the arithmetic mean, 1756 – 1781
4. Chance or design: Tests of significance
5. Theory of errors and methods of estimation
6. Fitting of equations to data
PART II: Inverse Probability by Bayes and Laplace, with comments on Later Developments
7. Introduction and probability: The philosofical background
8. Bayes, Price and the Essay, 1764 – 1765
9. Equiprobability, equipossibility, and inverse probability
10. Laplace's applications of the Principle of Inverse Probability in 1774
11. Laplace's general theory of inverse probability
12 The equiprobability model and the inverse probability model for games of chance
13 Laplace's methods of asymptotic expansion, 1781 and 1785
14. Lapace's analysis of binomially distributed observations
15. Laplace's theory of statistical prediction
16. Laplace's sample survey of the population of France and the distribution of the ratio estimator
PART III: The Normal Distribution, the Method of Least Squares, and the Central Limt Theorem. Gauss and Laplace, 1809 – 1828
17. Early history of the central limit theorem, 1810 – 1813
18. Derivations of the normal distributions as a law of error
19. Gauss's linear normal model and the methods of least squares, 1809 – 1811
20. Laplace's large-sample theory of linear estimation, 1811 – 1827
21. Gauss's theory of linear unbiased minimum variance estimation, 1823 – 1828
PART IV: Selected Topics in Estimation Theory, 1830 – 1930
22. On error and estimation theory, 1830 – 1890
23. Bienaymé's proof of the multivariate central limit theorem and his defense of Laplace's theory of linear estimation, 1852 – 1853
24. Cauchy's method for determining the number of terms to be included in the linear model and for estimating the parameters,1835 – 1853
25. Orthogonalisation and polynomial regression
26. Statistical laws in the social and biological sciences. Poisson, Quetelet, and Galton, 1830 – 1890
27. Fisher's theory of estimation, 1912 – 1935, and his immediate precursors
Readership: Statisticians, probabilists, actuaries, mathematicians, historians of science and advanced students
This is a marvallous book. It is a sequel to Hald's earlier volume, A History of Probability and Statistics and Their Applications Before 1750, although it includes a summary of the most important results from before 1750 and so can be read independently of that earlier volume. The original papers have been rewritten into a uniform modern terminology and notation, so that today's statiticians can read them with ease. The author comments that the book gives "a more detailed and extensive account than ever before of the development of statistical theory and its applications from 1750 – 1828", but that he realized the impossibility of continuing in that vein and so skipped several important contributions, restricting the material from 1830 to 1930 to a selection of topics. He also describes the book as a modern version of the classical works by Todhunter and Czuber.
Anyone with the slightest interest in the history of statistics, or in understanding how modern ideas have developed, will find this an invaluable resource.

Contents:
1. Introduction
2. Reckoning on death and chance with the Merchant's rule
3. Commercial currnets and first differences
4. The interplay of deception and accountability in the index numbers and moving averages of the Bank of England
5. Seasons, tides, and structures in cycle time
6. Laws of chance and error in the observation process
7. Laws of deviation and the capacity for shifting
8. A funny thing happened on the way to equilibrium: Economic and statistical ways of thinking around the turn of the century
9. Decomposition and functions of time
10. Autoregression, random disturbances, dangerous series, and stationary stochastic processes

This history starts with John Graunt's Bills of Mortality and ends with Herman Wold's A Study in the Analysis of Stationary Time Series. We are shown how Graunt, the haberdasher, used the algorithms of the earliest commercial arithmetics; how this was developed through the political arithmetic, descriptive statistics and statistical techniques of the next two centuries, to emergence of statistical theory in this century. It is a useful survey, although jargon sometimes gets in the way of understanding.
One thing, however, which has consistently come between me and the text is the quality of the editing, which varies from the slipshod to the appalling. In the text, typographical errors abound, names are misspell, and quotations miscopied; in the references, all foreign languages are treated badly – German is garbled, Italian is miscopied, French accents appear apparently at random and the transliteration of Russian is inconsistent.
I think that there is an interesting book here struggling to get out, but unfortunately the author made it hard work.

Contents:
1. What's life?
2. Before life began
3. The frozen accident
4. The oxygen menace
5. Artificial life
6. Flowers for Fibonacci
7. Morphobens and Mona Lisas
8. The peacock's tale
9. Walk on the wild side
10. An exaltation of boids
11. Reef wars
12. In search of secrets

This is the latest book from a well-known popularizer both of mathematics and of its relations with science. Stewart has produced many books conveying the excitement of the subject and also writes a regular column in Scientific American. This particular book is inspired by a brilliant tour de force On Growth and Form, by the zoologist D'Arcy Thompson, which first appeared in 1917, grew impressively to more than one thousand pages for its second edition in 1942 and is now available in abridged form (1961, Cambridge University Press).
The first secret of life is now the molecular structure of DNA. The whole thrust of this new book is the other secret, which is the way that mathematics underpins biological structure and behaviour. In particular life is a partnership between genes and mathematics. It is all very dazzling and inspiring, particularly for the general reader, in a field which is already itself experiencing rapid growth and is ripe for more!

Contents:
PART I: Dynamical Systems and Lotka-Volterra Equations
PART II: Game Dynamics and Replicator Equations
PART III: Permanence and Stability
PART IV: Population Genetics and Game Dynamics

Readership: Those interested in mathematical biology and/or biological mathematics

This book updates, restructures and replaces an earlier book by the same authors [The Theory of Evolution and Dynamical Systems, 1988, Cambridge University Press, reviewed in Short Book Reviews, Vol. 9, p. 10.]. The book is timely, since there has been explosive growth in the whole subject in the intervening period. The idea is to approach game theory and population dynamics ('Games of Life') as branches of dynamical systems theory, centrally making use of the replicator equations, which are essentially equivalent to generalized Lotka-Volterra equations. The modelling is deterministic, but the outcomes can, of course, be stochastic ('chaotic'). The approach is mathematical and the book provides excellent text for the subject of mathematical ecology. However, biologists would inevitably find the treatment tough going.

Contents:
1. Introduction
2. The analysis of segregation and population frequencies
3. The analysis of genetic linkage
4. The analysis of allelic associations
5. The analysis of continuous and quasi-continuous characters

Readership: Scientists engaged in human genetic research and advanced students of genetics and statistics

Increasingly sophisticated statistical methods are being used in human genetical research. The reader is assumed to have an understanding of basic statistical concepts but is gently introduced to the necessary knowledge of genetics. The emphasis is on understanding the genetic basis of many human diseases. The book brings together methods scattered over a wide literature. The assumptions underlying the methods of analysis presented are critically examined. The book can be recommended to those wishing to become familiar with this important area of work. New editions will be needed as the types of data available and methods required to analyze the data are advancing rapidly.

Table des Matières:
1. Le cheminement historique des plans d'expériences par J.-J. Droesbeke, J. Fine et G. Saporta
2. La planification des expériences et l analyse de la variance : Une introduction par P. Dagnelie
3. Les plans factoriels par A. Kobilinsky
4. Approche méthodologique des surfaces de réponse par D. Mathieu et R. Phan-Tan-Luu
5. Approche méthodologique des mélanges par D. Mathieu et R. Phan-Tan-Luu
6. L'approche Taguchi par D. Mathieu et R.Phan-Tan-Luu
7. Plans d'expériences optimaux pour modèles linéaires par J.-P. Gauchi
8. Plans d'expériences optimaux pour modèles de régression non linéaire par J.-P. Gauchi
9. Trois applications à l'entreprise par J.-P. Gauchi, G. Sapota, D. Mathieu et R. Phan-Tan-Lauu

Lecteurs : Ce livre s'adresse aux chercheurs, aux étudiants en statistiques des deuxième et troisième cycles

Cet ouvrage offre au lecteur un compte rendu détaillé et moderne des plans d'expérience. Bien qu'une variété d'auteurs aient traité différents aspects du sujet le style est assez cohérent et les éditeurs on adroitement réuni les différentes contributions.
Plans d'Expériences est un mélange subtil de théorie statistique et arguments combinatoires avec les jouant un rôle clé. Ce message est communiqué avec clarité bien que l'accent soit moins mis sur les propriétés combinatoires des plans d'expérience que sur l'analyse des données. Il y a un agréable mélange de théorie, conseils pratiques et exemples numériques. Il est particulièment plaisant de trouver des chapitres savants, qui cependant se lisent bien, sur les plans d'expérience optimaux (à la fors pour modèles linéaires et non linéaires), les mélanges et les méthodes de Taguchi. Ce n'est pas un livre pour quiconque est éstranger aux plans d'expérience ou à l'analyse de la variance mais il sera un outil de référence utile

Contents:
1. Introduction
2. CUSUM design
3. More about normal data
4. Other continuous distributions
5. Discrete data
6. Theoretical foundations about the CUSUM
7. Calibration and short runs
8. Multivariate data
9. Special topics
10. Software

Readership: Practitioners, students and teachers of quality methodologies, and anyone interested in the study of time-ordered data

This is a very nice monograph on the analysis of cumulative sums, accessible to any numerate reader with basic statistical training. It is a pleasant way to begin a personal study of the field. It can also be used as a supplementary text to broaden traditional courses in time series or quality control. Readers seeking more depth will find apt guidance in the authors' comments at the end of each chapter and in the well-chosen reference list at the end of the book. The text is supported by sets of data and software that can be downloaded from the website of the School of Statistics of the University of Minnesota, a nice bonus to this already friendly book.

Contents:
1. Highlights
2. Concepts
3. Quick reference
4. Computations in S
5. Objects, databases, and chapters
6. Creating functions
7. Creating classes
8. Creating methods
9. Documentation
10. Connections
11. Interfaces to C and Fortran

Readership: Current and future users of S and S-Plus; students, teachers and practitioners of data analysis; computer scientists; those with an interest in statistical computing

In the ten or so years since we discovered S, we have come to rely on its natural expression of data-analytic ideas as well as its power and versatility. This authoritative book, based on a new version of the language, takes the symbiosis of data and programming language to new levels of expression.
The writing is accessible, indeed encouraging; the ideas are presented with exceptional clarity. The reader is taken initially through a sophisticated example, given an overview of the main features of the S language, and then is led more deeply into concepts that reveal the full power of objects, classes and methods for data analysis.
Throughout are thoughtful comments on how to access the ideas most effectively; for example, abstracts at the beginning of each chapter convey not only the flavour of the contents but how the chapter complements the reading of other chapters. As a welcome and unadvertised bonus, we find numerous insights into how a master programmer thinks about programming.

Contents:
1. Introduction to experimental design
2. Informal analysis and checking assumptions
3. Formal anova: Decomposing the data and measuring variability, testing hypothesis and estimating true differences
4. Decisions about the content of an experiment
5. Randomization and the basic factorial design
6. Interaction and the principle of factorial crossing
7. The principle of blocking
8. Working with the four basic designs
9. Extending the basic designs by factorial crossing
10. Decomposing a data set
11. Comparisons, contrasts, and confidence intervals
12. The Fisher assumptions and how to check them
13. Other experimental designs and models
14. Continuous carriers: Visual approach to regression, correlations and analysis of variance
15. Sampling distributions and the role of the assumptions

Readership: Students with no previous background in statistics who wish to design and analyze experiments

This is a very long book for the material covered simply because the author has a gentle, conversational style and enjoys discussing the various facets of experiments at some length. If you pursue self-study with this book as your guide, you will be extremely happy with it, particularly if you are not in a hurry. A reference book it is not, however. For example, look up "transformation, diagnostic plot for" in the index and you go to p. 489 and see Kelly's enzyme data. No Kelly, no enzyme in the index. Turning a few pages, however, gives you Kelly's hamsters on p. 485, and there you see a reference to Section 2 of Chapter 1. From page 13 you see that Kelly is actually earlier; indeed she opens Chapter 1. As a colleague of the author remarked (p. xx) "I tell the students it's all there, but they really have to read it." The same advice would apply to an instructor who wanted to use the book as a class text. It is an excellent elementary text on experimental design and analysis, particularly for self-study, written with love of the subject. Will you like it? Depends on you!

Contents:
1. Nature of exact and optimum tests in mixed linear models
2. Balanced random and mixed models
3. Measures of data imbalance
4. Unbalanced one-way and two-way random models
5. Random models with unequal cell frequencies in the last stage
6. Tests in unbalanced mixed models
7. Recovery of inter-block information
8. Split-plot designs under mixed and random models
9. Testing using generalized P-values
10. Multivariate mixed and random models

This book is concerned with hypothesis testing, in contrast to point or interval estimation and prediction, in balanced and unbalanced mixed effects models, and brings together many recently developed results in the area of exact and optimum inference for variance components models.
The first chapter provides an introduction to the extension of Wald's variance component tests to the more general framework, where the model is expressed in terms of a set of primary effects (fixed or random), a set of fixed effects and a set of random effects, and then discusses the nature of exact and optimum tests of hypotheses concerning these primary effects.
In Chapter 2, which deals with balanced data, it is shown that the usual F-tests obtained from the analysis of variance are optimal. The effects of various types of imbalance are introduced in the next chapter and a measure of imbalance is developed to provide a better understanding of the effects of imbalance on the efficiencies of estimators and tests. Use of this measure is illustrated with a variety of different models including one with a mixture of cross-classified and nested effects.
Chapters 4,5 and 6 deal with the development of exact, optimal and approximate tests for different models ranging from a one-way random effects model to a general unbalanced mixed model.

Contents:
1. Introduction
2. Some basic ideal of smoothing
3. Statistical properties of smoothers
4. Data-driven choice of smoothing parameters
5. Classical lack-of-fit tests
6. Lack-of-fit tests based on linear smoothers
7. Testing for association via automated order
8. Data-driven lack-of-fit tests for general parametric models
9. Extending the scope of application
10. Some examples

Readership: Postgraduate students and statisticians interested in data smoothing

The book relies heavily upon what is usually referenced as smoothing or nonparametric methods. These methods are based on various approximations in the relative co-ordinate system that is related to, or centred at, the point of interest. In the classical approximation, the analysis, regression analysis in statistical settings, is performed in an absolute co-ordinate system that is not explicitly related to the point of interest. The first four chapters give an extensive survey of smoothing methods. Most of them can hardly be called nonparametric because they actively use such 'smoothing' parameters as the band-width, the kernel order, expansion with respect to various basic functions, etc. The rest of the book is almost entirely devoted to the lack-of-fit tests for the traditional regression models which use the results of smoothing as an alterative to the zero hypothesis.

Contents:
PART I: Conventional Simulation
1. Systems, models, and simulation
2. Random numbers, variates, and stochastic processes generation
3. Output analysis of discrete-event systems via simulation
4. Variance reduction techniques
PART II: Modern Simulation
5. Sensitivity analysis and optimization of discrete-event static systems (DESS)
6. Sensitivity analysis and optimization of discrete-event static systems dynamic systems: Distribution parameters
7. Sensitivity analysis and optimization of discrete-event static systems dynamic systems: Structural parameters
8. Response surface methodology via the score function method
9. Estimating rare-event probabilities and related optimization issues

Readership: Students in statistics and probability theory; students and researchers interested in stochastic simulation

The book has a textbook format and contains a sufficient number of examples and exercises. However, the second part, namely Chapters 5-9, gravitates more to a monograph style and may attract a wider audience, including engineers and scientists, who apply stochastic discrete models or are interested in the Monte-Carlo-type calculations. Adding the words "stochastic and discrete-event (simulation and modelling)" to the title would probably give a better image of what is discussed in the book.

Contents:
1. Introduction
2. Single-index models
3. Binary response models
4. Deconvolution problems
5. Transformation models

Readership: Graduate students and applied researchers familiar with econometrics theory at the level taught in graduate-level courses in leading universities

Models and estimation problems that involve an unknown function and an unknown finite-dimensional parameter are called semiparametric. In contrast models that include unknown functions and no finite-dimensional parameters are called nonparametric. This book is concerned with estimation of semiparametric models. It emphasizes ideas rather than technical details; for example, it gives heuristic explanations of how important results are proved rather than formal proofs. Each chapter contains a real-data application as well as examples without data of applied problems in which semiparametric methods can be useful. The discussion is entirely classical and asymptotic in nature. Individuals with small samples or Bayesian sympathies will feel short-changed, but most readers of econometrics will find this book to be valuable introduction to the extensive theoretical literature permeating econometrics journals.

Contents:
1. An introduction to survey participation
2. A conceptual framework for survey paticipation
3. Data resources for testing theories of survey participation
4. Influences on the likelihood of contract
5. Influences of household characteristics on survey cooperation
6. Social environmental influences on survey participation
7. Influences of the interview
8. When interviewers meet householders: The nature of initial interactions
9. Influences of householder-interviewer interactions on survey cooperation
10. How survey design features affect participation
11. Practical survey design acknowledging nonresponse

This book is concerned with the statistical and practical aspects of non-response in sample surveys, and with the features of the survey design that might affect the participation rates. Although the first chapter reviews the statistical properties of survey estimates subject to non-response, the majority of the text considers the practical and conceptual issues which might influence non-response when carrying out a survey. The non-response phenomenon is dissected into its several components; non-contact, refusal and non-availability of the selected household unit. The focus is on understanding the reasons for non-response, particularly the reasons for respondents refusing to take part. The book provides a comprehensive investigation of the factors which influence a respondent's decision to participate in a survey, and it is hoped that a fuller appreciation of the effects of these factors will enable researchers to avoid those that adversely influence the participation rate. The factors discussed range from environmental characteristics to the initial interaction between interviewer and house-holder. The ideas pr5esented are illustrated with empirical analyses based on eight different household surveys, some matched with census recods and others with interviewer observations.
The reasons for non-response are many and evidently depend on the particular features of the survey as well as on the survey design itself. This book gives a detailed discussion of the many influences involved and provides a framework for a deeper understanding of non-response in household surveys.

Contents:
1. The nature of multivariate data analysis
2. Vector and matrix operations for multivariate analysis
3. Vector and matrix concepts from a geometric viewpoint
4. Linear transformations from a geometric viewpoint
5. Decomposition of matrix transformations: eigenstructures and quadratic forms
6. Applying the tools to multivariate data
APPENDIX A: Symbolic Differentiation and Optimization of Multivariable Functions
APPENDIX B: Linear Equations and General Inverses

Readership: Experimental scientists, statisticians, students of multivariate analysis

This book updates the earlier edition published in 1976. It provides a careful and thorough introduction to vectors and matrices. Especially valuable is the material providing geometric interpretations. The applications of Chapter 6 concentrate on multiple regression, multiple discriminant analysis and factor analysis. Each chapter ends with a summary and sets of numerical exercises, all of which are provided with solutions. A particular strength of the book is the frequent use of small numerical examples which, for example, actually demonstrate the useful properties of determinants, and make absolutely clear what is meant by operations like the multiplication of matrices. The book is designed for readers who have no prior knowledge of matrix theory, and specifically for students in the behavioural and administrative sciences. However, it is also very clear and useful that it has material of value to anyone using multivariate methods. It should be on the reading list for all courses on multivariate analysis.

Contents:
1. One-sample problems
2. Two-sample problems
3. Linear models
4. Experimental designs
5. Bounded influence and high-breakdown methods
6. Multivariate modes
APPENDIX: Asymptotic Results

This book provides a unified methodology for nonparametric methods. It is particularly interesting to observe the direct analogies which can be made with traditional least-squares methodology, with a L1 norm taking the place of Euclidean distance. The book is inherently mathematical in its approach, making consistent use of ideas from geometry. However, by placing the asymptotic theory in a forty-three page appendix and providing many detailed illustrations using real data, the authors have ensured that the material is accessible and full of interest. Details of how to obtain the RGLM (Robust General Linear Models) computer package used in the book are available on the world wide web. It is suggested that material from the first four chapters could form the basis of a graduate course in rank-based methods and the large number of exercises will assist in this. This volume is destined to become a standard source of reference for rank-based statistical methods.

Contents:
1. General concepts
2. Introduction to multivariate statistical procedures
3. Correlation
4. Factor analysis
5. Canonical correlation
6. Multiple regression
7. Multivariate analysis of variance
8. Multivariate analysis of covariance
9. Principal components
10. Multiple discriminant analysis
11. Cluster analysis
12. Multiple logic regression
13. Coefficient of variation
14. Correspondence analysis
15. Multivariate probit analysis
16. Multivariate time series modelling
17. Multivariate spatial measures
18. Multivariate data preparation and plotting
19. Summary and generalizations of multivariate quantitative procedures

This book aims to introduce researchers in the geosciences to a wide range of multivariate techniques. The techniques are grouped according to whether they focus on variables or individuals, and further subdivided according to the type of measurements they can handle. Each chapter deals with a particular technique, typically in a dozen pages or less, by presenting the concepts, some definitions, an overview of the methods, some selected technical aspects, and a few numerical examples taken from the scientific literature.
The idea behind the book is a good one, but I wonder what a geoscientist will make of it. From a statistician's perspective, the descriptions of the techniques are so brief as to be confusing (indeed the chapter on multivariate probit analysis lacks a description altogether), technical aspects are discussed without prior foundation, explanations are sometimes misleading, notation pops up without warning and without consistency between chapters, the numerical examples are presented straight out of the source texts without any discussion or interpretation, and even a casual reading reveals obvious typographical errors. Not a book I would recommend.

Contents:
1. Introduction
2. Overview of contextual models
3. Varying and random coefficient models
4. Analyses
5. Frequently asked questions

Readership: Researchers, lecturers and students dealing with the statistical analysis of hierarchical data

This book "for researchers and students in the social sciences with no strong background in statistics and linear algebra… written with practitioners in mind" aims to give a guide to the principles and practicalities of multilevel modelling, keeping formal mathematics to a minimum. Essentially the discussion throughout is in the context of two-level data with normally distributed responses and with homoscedesticity at level one.
One-third of the book (Chapter 4) is devoted to a range of analyses, using the MLn package, of an actual educational two-level set of data (mathematics scores of five-hundred and nineteen students in twenty-three schools, with seven explanatory variables). No less than twenty different models are discussed and compared.
Chapter 5 'Frequently asked questions' addresses some important questions asked by researchers subscribing on e-mail to the multilevel mailing list.
Although not without some errors, the book provides a clearly written introduction to the basics of multilevel analysis. For someone with substantial data to analylze, it would probably be best used in conjunction with a more extensive (while equally reader-friendly) text such as I. Plewis, Statistics in Education, Arnold, 1997, or G. Woodhouse, A Guide to MLn for New Users, Institute of Education, University of London, 1995.

Contents:
1. Basic concepts
2. Inferential problems in estimating a finite population total
3. Direct, synthetic and composite estimators
4. Bayes and empirical Bayes estimation under random effects models
5. Ramifications of Bayes estimation procedure
6. Best linear unbiased prediction under mixed effects models
7. Bayes, empirical Bayes ahd hierarchical Bayes
8. Small area estimation of population counts
9. Small area estimation using time series methods

The book contains an extensive collection of statistical techniques, that can be used in analysis of data measured/observed at some specified units ('small areas' in the author's terminology). The most specific of them are related either to sampling from finite populations or to mixed effects models. Others are less specific and include best linear unbiased estimation/prediction, the Kalman filter technique and the Bayesian procedures. To read the book one must be familiar with some solid university course in multivariate statistics.

Contents:
1. Introduction to stereology
2. The co-area formula
3. Rotation invariant measures on Lpn
4. The classical Blaschke-Petkantschin formula
5. The generalized Blaschke-Petkantschin formula
6. Local slice formulae
7. Design and implementation of local stereological experiments
8. The model-based approach
9. Perspectives and future trends
APPENDIX: Invariant measure theory

Readership: Researchers, teachers and graduate students in mathematical statistics and probability; specialists in stereology, integral geometry and geometric measure theory

Stereology is the spatial sampling theory arising from measurements made on microscope samples. The theory guarantees the validity of three-dimensional interpretation. Local stereology came out of the practical needs of microscopists following developments in observational techniques. Inference is to be derived from sections being taken through reference points such as cell nuclei. This theory, dating only from the 1980s, involves advanced mathematics, in particular geometric measure theory, where the sections are d-dimensional affine subspaces Rn. The co-area formula is the mathematical tool. This is a version of Fubini's theorem involving generalized Jacobians for transforming integrals with respect to Hausdorff measures. The book is thus aimed at mathematicians who are sought out for advice by microscopists. Chapter 7 shows how the methods may be applied in practice, and cites their implementation by stereologists. The book ends with a valuable eleven-page list of references, ranging from "Six billion neurons lost in AIDS" to "A kineamatic formula and moment measures of random sets".
This book adds weighty theoretical support to the increasingly important field of spatial sampling.

Contents:
PART I: Historical Overview
1. Biography of Jaroslav Hájek
2. Essays
3. Publications of Jaroslav Hájek
4. Hájek's PhD students
PART II: Collected Works of Jaroslav Hájek

Jaroslav Hájek died in 1974 at the age of 48. The year before his untimely death, he started the Prague Symposia on Asymptotic Statistics, which since then have been organized every five years and are considered to be very important scientific events. It is interesting to see that twenty-five years later, Hájek's thoughts and writings on asymptotic estimation theory are still very influential, not only in the famous Department of Probability and Statistics of Charles University in Prague, but also in many other statistics research groups world wide. Many of us will of course have knowledge of his famous book on Theory of Rank Tests, written with Z. Sidák in 1967. Hájek also published fundamental results and methods in statistical inference in stochastic processes, in finite population sampling and in the asymptotics of parameter estimators.
M. Hušková, R. Beran and V. Dupaè have compiled (almost) all of his articles written in the period 1949-1974. Several of the early papers in Czech now appear for the first time in English translation. There are also six essays by some of his famous colleagues and former students, who write on the personality of Hájek and on his inspiration and influence for modern statistics. The whole collection is a valuable contribution to the statistical community. It is a unified body of important material which is highly recommended reading for researchers in statistical inference.

Table des Matières :
1. Introduction générale
2. La collecte des données
3. La statistique descriptive à une dimension
4. La statistique descriptive à deux dimensions
5. La probabilité mathématique et les distributions théoriques : généralités
6. Les principales distributions théoriques à une dimension
7. Les distributions théoriques à deux dimensions
8. Les distributions d'échantillonnage
9. Les problèmes d'estimation
10. Les tests d'hypothèses

Lecteurs : Etudiants et enseignants dans le domaine de l'agronomie et de la biologie

Selon l'auteur, la statistique est l'ensemble des méthodes qui permettent de rassembler et d'analyser les données numériques. En particulier les méthodes sont utilisables dans l'agronomie, qui est le domaine d'intérêt principal de ce livre. Les deux volumes de 'Théorie et méthodes statistiques : applications agronomiques' (1969-1970) sont devenus en presque trente ans des ouvrages classiques en langue française. Ces livres ont connu plusieurs révisions et le présent Tome 1 constitue encore une nouvelle remise à jour. La théorie statistique classique est présentée avec un minimum de mathématiques, ce qui donne ouverture à un grand nombre d'étudiants. Il est intéressant de voir que les chapitres se terminent par les énoncés d'exercises, dont les solutions se trouvent à la fin. Ce livre peut donc bien servir comme manuel dans l'enseignement du premier cycle.

Contents:
PART I: Prologue and Preliminaries
PART II: Singularly Perturbed Markov Chains
PART III: Control and Numerical Methods

Readership: Graduate students and researchers in applied probability and operations research, engineers with interest in stochastic modelling

In practical applications of stochastic models, one is often faced with a degree of complexity that makes a standard textbook analysis impossible. In such cases it often helps to relate different parts of the model of different time scales, and an expansion with respect to faster oscillating factors might considerably reduce the complexity and enable an analysis of the quantities of interest. The book explains such techniques for continuous-time Markov chains and gives several applications. The mathematical level is rigorous, but accessible. The methods explained in this book deserve to be more widely known.

Contents:
1. Introduction: Applications and issues
2. Applications to learning, state dependent noise, and queueing
3. Applications to signal processing and adaptive control
4. Mathematical background
5. Convergence with probability one: Martingale difference noise
6. Convergence with probability one: Correlated noise
7. Weak convergence methods for general algorithms
8. Weak convergence methods for general algorithms
9. Applications: Proofs of convergence
10. Rate of convergence
11. Averaging of the iterates
12. Distributed/Decentralized asynchronous algorithms

Readership: Statisticians and engineers interested in "online" identification and control of stochastic systems

As befits its title, this book is concerned with stochastic approximation as it has evolved since its beginning in 1951 to deal with problems of identification and control of stochastic systems. The first three chapters present a number of examples of potential applications that serve to motivate the theoretical considerations that follow in Chapters 4 to 12. The emphasis is on convergence theorems involving convergence with probability one or in distribution. The authors return only briefly to discussion of the applications in view of the theoretical developments. The senior author has written numerous papers and a previous book on the subject, so not surprisingly the present book is focused on his many contributions. While this book contains a large number of interesting specialized results, I found it easier to get a broad overview of the subject from M. Duflo's Random Iterative Models [Short Book Reviews, Vol. 17, p. 48], which, however, is concerned almost exclusively with almost sure convergence.

Contents:
1. Introduction and discrete-time models
2. The continuous-time market model
3. The continuous-time portfolio problems
4. Constrained continuous-time problems
5. Portfolio optimisation in the presence of transaction costs
6. Non-utility based portfolio selection models

The theory of portfolio optimization was founded by Harry Markowitz in 1952. His essentially one-period optimization problem was based on a mean (return) and standard deviation (risk) representation, leading to the well-known construction of the efficient investment frontier. Modern continuous-time stochastic calculus yields two basic approaches to the continuous-time optimization problem: stochastic control theory and the martingale approach. The author compares and contrasts these alternative approaches in a very clear way. Various examples, historical comments, extended remarks on how and why, keep the attentive reader on the right track. Consequently, the book yields an excellent text for a graduate course on the subject.

Contents:
1. Introduction. A model-based approach to actuarial science
2. Loss distributions. Models for the amount of a single payment
3. Frequency distributions. Models for the number of payments
4. Aggregate loss models
5. Credibility theory
6. Long-term models

Readership: Students of actuarial science and risk theory; professional reference for insurance and risk management

This book aims to show how the process by which funds flow into and out of an insurance system may be modelled. It requires a background in probability and mathematical statistics. Many of the exercises are taken from qualifying examinations on loss distributions, credibility and risk theory for actuaries. Within the text are sections on Bayesian estimation, empirical Bayes parameter estimation and Brownian motion. There are appendices on inventories of probability distributions, optimization, discretization and simulation.
The resulting book is an eye-opener for this reviewer. Statisticians often feel that the statistics training given to candidates for professional qualifications in finance, engineering, etc. leaves them inadequately prepared for the interpretation of complex data, which should be reserved for the professional statistician. This would not apply to actuaries trained from this book. It provides a model of how statistics might be taught. This book is worthy of classic status, and should be considered for adoption as the standard for actuarial studies.

Contents:
Introduction
1. The finite population exponential model
2. Lifetime data analysis
3. Counting the number of failures
4. Strength of materials
5. The economics of maintenance and inspection
6. Network reliability
7. System failure analysis: Fault trees
8. System availability and maintainability
9. Influence diagrams
10. Making decisions using influence diagrams
APPENDIX A: Classical Statistics is Logically Untenable
APPENDIX B: Bayesian Decision Analysis is Self-Consistent

This is an idiosyncratic but interesting discussion of engineering reliability and maintenance. Parameters of interest are defined whenever possible in the context of finite populations, and probability models are developed through symmetry or 'indifference' conditions. 'Infinite' population models arise as limiting distributions. System reliability and decision analysis are emphasized. It should be clear from the titles of the two appendices what approach is taken to inference and decision. This book does not cover the field as well as others such as Høyland and Rausand's System Reliability Theory [Short Book Reviews, Vol. 15, p. 44.] and references to the literature are scant. Nevertheless, it can be recommended as a provocative presentation of the author's view of reliability.

Contents:
1. Control in a non-stationary world
2. Control charts for frequencies and populations
3. Control charts for measurement data
4. Modelling process dynamics and forecasting using exponential smoothing
5. Time series for process disturbances
6. Process adjustment using feedback control: Manual adjustment charts
7. Control of a process with inertia
8. Feedback control when there are adjustment costs
9. Including the costs of surveillance: How often should you sample?
10. Direct process monitoring and cuscore charts: Looking for signals in noise
11. Simultaneous adjustment and monitoring
12. A brief review of time series analysis

Readership: Those working in statistical process control who want to learn a little about engineering process control

This book is aimed at building on traditional ideas of statistical process control to incorporate notions of feedback control. The treatment is at a very elementary level using graphical methods wherever possible and totally avoiding the use of calculus. For the reviewer's taste the overall level is a bit too elementary. However, the book would be useful to practitioners of Statistical Process Control who are looking to add some feedback control to their repartoire.

Contents:
1. Introduction to the theory of probability metrics
2. The nature of limit theorems
3. The normalization of random sequences
4. Centres and scatters of random variables
5. Classical theory of limit theorems for sums of independent random variables
6. Generalizations of the classical theory of summation of independent random variables

The classical theory of limit theorems for sums of independent random variables is presented, together with its more recent generalizations and refinements. Attention is given not only to the limit theorems, but also to the rates of convergence, which are described with the help of various metrics. The historical remarks and the extensive reflections on theory and methods make the book particularly interesting.

Contents:
1. Basic notions
2. Semi-martingales
3. Inequalities
4. Laws of large numbers
5. Central limit theorem
6. Convergence of semi-martingales
7. Rate of weak convergence
8. Applications of martingales

The main body of the book deals with the important limit theorems, law of large numbers, central limit theorem, functional central limit theorem, together with results on the rate of convergence. The primary notions on martingales and semi-martingales are introduced in the first chapters. In the last chapter, the power of martingales is illustrated in the domains of U-statistics, rank statistics, statistics of exchangeable variables and stochastic exponential statistics. There are interesting bibliographical notes and an impressive list of about four-hundred references to the literature.

Contents:
1. Risk theory
2. Fluctuations of sums
3. Fluctuations of maxima
4. Fluctuations of upper order statistics
5. An approach to extremes via point processes
6. Statistical methods for extremal events
7. Time series analysis for heavy-tailed processes
8. Special topics

Readership: Financial engineers, actuarial students, professional actuaries, undergraduate and graduate students interested in insurance or mathematical finance, probabilists, statisticians

Extreme events play a central role in the (re)insurance industry and are increasingly important throughout the financial services industry with the current interest in credit insurance, mortgage-backed securities, catastrophic insurance futures, etc. On the other hand there is a rich probabilistic theory of extreme values with the classical characterization of the Fréchet, Weibull and Gumbel extreme value distributions and their domains of attraction as well as ruin theory, fluctuation theory and time series analysis. This book is written very much with a wide range of readers in mind and presents the essential aspects of these probabilistic theories in an easily accessible style. However the book contains a wealth of information and can be read at many levels. For example it includes a chapter introducing the approach to extreme value theory via point processes including the methodology of weak convergence of random measures. It also contains an introduction to exploratory data analysis for extremes and the statistical estimation theory for the GEV (generalized extreme value distribution). An-other important topic is time series analysis for heavy-tailed processes. The final chapter introduces a host of special and specified examples from the insurance and finance industries. For example, here you can find an introduction to extremes of ARCH and GARCH processes, a detailed introduction to reinsurance treaties, stable processes and self-similarity. The book also includes appendices on some basic mathematical topics which may not be generally familiar to all the intended readers and an extensive set of six hundred and forty-six references. This book will undoubtedly become a standard reference book for the insurance and finance communities.

Contents:
1. Introduction
2. The predictability of asset returns
3. Market microstructure
4. Event-study analysis
5. The capital asset pricing model
6. Multifactor pricing models
7. Present-value relations
8. Intertemporal equilibrium models
9. Derivative pricing models
10. Fixed-income securities
11. Term-structure models
12. Nonlinearities in financial data

Readership: Econometricians, mathematicians interested in financial markets, economists and finance experts

What are the facts, methods and techniques underlying the intersection of economics, finance and statistics? The authors aim at giving a comprehensive answer to this question. By definition of the project, one is likely to be criticized by specialists in each of the larger fields: for instance, the professional probabilist may express doubts on the depth of treatment of Brownian motion and Itô calculus. The economist on the other hand would have liked a greater emphasis on fundamentals. Finally, the professional statistician may perhaps press for a more critical discussion on certain tests used. This may all be true, but the fact of the matter is that in the real world compromises have to be made. In my view, the authors have struck the right level. They clearly are competent in the larger field, but they strive at reaching a level of presentation accessible to a much wider audience. The professional statistician will no doubt learn a lot of the real economic questions underlying financial markets. On the other hand, the level of detail on the underlying stochastic methodology is sufficient to satisfy both the mathematician as well as the economist. The book covers a remarkable range of highly important topics. The about eight hundred references give a very solid basis for further reading. I consider the book a must for the intended readership.

Contents:
PART I : Spot and Futures Markets
1. An introduction to financial derivatives
2. The Cox-Ross-Rubinstein model
3. Finite security markets
4. Market imperfections
5. The Black-Scholes model
6. Modifications of the Black-Scholes model
7. Foreign market derivatives
8. American options
9. Exotic options
10. Continuous-time security markets
PART II: Fixed-Income Markets
11. Interest rates and related contracts
12. Models of the short-term rate
13. Models of instantaneous forward rates
14. Models of bond prices and LIBOR rates
15. Option valuation in Gaussian models
16. Swap derivatives
17. Cross-currency derivatives
PART III: Appendices

Readership: All those who want to understand how mathematics and finance really go together

In the Preface, the authors state that: "Our hope is that this book will help to bring the mathematical and financial communities together, by introducing mathematicians to some problems arising in the theory and practice of financial markets, and by providing finance professionals with a set of useful mathematical tools in a comprehensive and self-contained manner." To a large extent, the authors have been successful in doing just that. The book contains a wealth of material expressed in a clear mathematical way. A definite bonus is the very extensive list of references which gives the reader a most welcome basis from which to explore further the realm of mathematical finance. Especially Part II on Fixed Income markets will please many students and experts alike. The book can be used ideally for both an introductory as well as an advanced text on mathematical finance. Though almost every week, we see a new text on the subject, this one stands out both in quality and coverage. It is a must for any library (private or public) which has mathematical finance as a subsection. I very much enjoy browsing through it.

Contents:
Volume I: Theory
1. Introduction
2. The Monge-Kantorovich problem
3. Explicit results for the Monge-Kantorovich problems
4. Duality theory for mass transfer problems
5. Applications of the duality theory
6. Mass transhipment problems and ideal metrics
Volume II: Applications
7. Relaxed or additional constraints
8. Probabilistic-type limit theorems
9. Mass transportation problems and recursive stochastic differential equations and empirical measures

In over nine hundred pages of tightly written material, the authors present a comprehensive overview of numerous problems, results and applications that have to do with distances between probability measures. In their preface to Volume I, the type of problem is traced back to 1781 when Gaspard Monge posed the question of finding optimal procedures for transporting soil from one area to another. Today the prototypical theoretical description of such a problem is described in terms of optimising of such a problem is described in terms of optimising over joint distributions with specified marginals. More specifically, the two central problems are (1) the transportation problem of Monge and Kantorovich which focuses upon minimization of E{c(X,Y)} for a given cost of function c, the infimum being over all joint laws for X and Y; and (2) the transhipment problem of Kantorovich and Rubinstein in which only the difference of the marginal laws of X and Y is specified. The general duality theory for these problems is presented along with specific optimality results. Applications are numerous: the authors list several of these with references, including econometrics, quality control and reliability. Most of the coverage in Volume II on Applications is on the theoretical modifications suggested by these applications, as well as upon the applications to theoretical areas of probability and algorithms. The former include interesting insights into central limit theorems and rates of convergence.
The authors are to be commended for their extensive list of five hundred and ninety-six references (after removing two that are listed twice) and for an exemplary index of symbols covering eleven pages. Although a different title for this two-volume treatise might identify its coverage more clearly, the volumes present a valuable contribution, comprising as they do a scholarly and thorough presentation of the theory and broad applicability of metrics for probability measures.

Contents:
1. Fractal measures
2. Integrals
3. Integrals and fractals
4. Probability
5. Probability and fractals

The title and chapter headings clearly indicate the trichotomous outline of this monograph. Because the book is of modest length, its coverage of each topic is limited. Though the presentation is necessarily terse, with results and remarks given in staccato precision, the author succeeds in making the text quite readable. There are numerous examples, exercises and historical notes, as well as over two hundred and seventy references. The last chapter on random fractals is especially well written. The exercises vary greatly in difficulty. The caveat in the Preface that "(t)hey should always be understood in the sense of 'prove or disprove'…" is disconcerting, especially for student readers who are asked, for example, in the first exercise of Chapter 4 to 'prove' the existence of a countably additive extension of a finitely additive probability defined only on an algebra. Overall, the author provides us with valuable overviews of the foundations of three linked areas of current importance.

Contents:
1. Introduction
PART I : Theory and Basics
2. A quick course in Bayesian statistics and decision theory
3. New advances in numerical Bayesian techniques
PART II : Applications in Econometrics
4. Imposing economic theory
5. Studying parameters of interest
6. Unit root and cointegration tests
7. Model specification uncertainty
8. Forecasting
9. More realistic models through numerical methods
PART III: Applications to Economic Decision Making
10. Decision theory applications

Readership: Advanced undergraduates and up; no priorknowledge of Bayesian statistics nor computer simulation methods is assumed

The two goals of this terse little book are to convince readers of the usefulness and inherent advantages of Bayesian statistics relative to sampling theory, and to provide a road map of applied economic questions that can be answered empirically with Bayesian methods. The book definitely achieves its second goal, and may even achieve its first goal among classical researchers who read it with an open mind. Researchers in economics who are curious about the Bayesian approach, but who have no formal training, will find this book's non-theoretical, empirical approach most appealing. Students looking to start Bayesian empirical projects will find the concise treatment of many applications as an efficient in-road into more advanced analyses contained in the numerous references. Gibbs sampling is discussed at an elementary level, but more advanced MCMC techniques are not. Bayesian re-searchers are likely to find the discussion very elementary, but they are not the primary intended audience. In summary, this book helps fill a gap in the often neglected low-end range of Bayesian pedagogy. It is a most welcomed addition.

Contents:
1. Introduction
2. The mixed Poisson distribution
3. Some basic concepts
4. The mixed Poisson process
5. Some related processes
6. Characterization of mixed Poisson processes
7. Reliability properties of mixed Poisson processes
8. Compound mixed Poisson distributions
9. The risk business

Motivated mainly by examples stemming from actuarial risk theory, the author gives a thorough review (mainly) of analytic properties of mixed Poisson processes. The latter are (conditional) Poisson pro-cesses for which the intensity follows a given distribution function. The various possible definitions of such processes are given, together with their proper-ties and numerous examples. The book is very carefully written and will give the student as well as the re-searcher in actuarial mathematics a reliable source on this important class of claim arrival processes. The book will of course also be of interest to the general applied probabilist.

Contents:
1. Introduction
2. The general framework
3. An alternative approach: E-sufficiency
4. Asymptotic confidence zones of minumum size
5. Asymptotic quasi-likelihood
6. Combining estimating functions
7. Projected quasi-likelihood
8. By-passing the likelihood
9. Hypothesis testing
10. Infinite dimensional problems
11. Miscellaneous applications
12. Consistency and asymptotic normality for estimating functions
13. Complements and strategies for application

Readership: Statistical scientists and graduate students interested in parameter estimation

In Heyde's terminology, a "quasi-score" is an optimal zero-mean estimating function in a specified class. In many cases the class consists of combinations of specified elementary estimating functions. The book outlines a general framework for the finite sample and asymptotic theory of the optimal combinations, and addresses applications to generalized linear models, GEE (Generalizing Estimating Equations) theory, and inference for stochastic processes.
The goal of the development is to provide a robust and flexible "semi-parametric" alternative to likelihood. In the course of examining quasi-likelihood analogues of likelihood methods, the book gathers together for the first time much of the associated theory developed by the author and others over the last fifteen years. Thus the book is useful particularly as a survey and synthesis of asymptotic theory and optimality results. The treatments of confidence zone construction and of the numerical solution of estimating equations are relatively brief. However, the treatment overall is straight-forward and comprehensive, with numerous and wide ranging illustrations, and is thus very suitable as an introduction to the subject as well as a reference for ongoing study.

Contents:
1. Elements of measure theory
2. Processes, distributions, and independence
3. Random sequences, series, and averages
4. Characteristic functions and classical limit theorems
5. Conditioning and disintegration
6. Martingales and optional times
7. Markov processes and discrete-time chains
8. Random walks and renewal theory
9. Stationary processes and ergodic theory
10. Poisson and pure jump-type Markov processes
11. Gaussian processes and Brownian motion
12. Skohorod embedding and invariance principles
13. Independent increments and infinite divisibility
14. Convergence of random processes, measures, and sets
15. Stochastic integrals and quadratic variation
16. Continuous martingales and Brownian motion
17. Feller processes and semigroups
18. Stochastic differential equations and martingale problems
19. Local time, excursions, and additive functionals
20. One-dimensional SDEs and diffusions
21. PDE-connections and potential theory
22. Predictability, compensation, and excessive functions
23. Semimartingales and general stochastic integration

Readership: Mathematicians, students and researchers alike, wanting a broad overview of modern probability theory

I am often asked by mathematicians, nonprobabilists, for literature on "a broad introduction to modern stochastics". Recently this question comes from many wanting to continue afterwards with a serious study of mathematical finance. Due to this book, my task for answering is made easier. This is it! A concise, broad overview of the main results and techniques of more advanced stochastic methods. From the table of contents, it is difficult to believe that behind all these topics a streamlined, readable text is at all possible. It is; convince yourself. I have no doubt that this text will become a classic. Its main feature of keeping the whole area of probability together and presenting a general overview is a real success. Scores of students (under-graduates, graduates) and indeed researchers will be most grateful!

Contents:
1. Why multivariate statistics?
2. Joint distribution of several random variables
3. The multivariate normal distribution
4. Parameter estimation
5. Discrimination and classification, round 1
6. Statistical inference for means
7. Discrimination and classification, round 2
8. Linear principal component analysis
9. Normal mixtures

The aim of this long book is to integrate theoretical and practical aspects of multivariate analysis. The practical emphasis comes early, with a short introduction to data. Some data tables span eight pages, but data and programs in S-plus and GAUSS are available by ftp. This is then followed by some two hundred and fifty pages of theory, prior to the first multivariate techniques. Theory is in standard Theorem/ Lemma/Proof format. In contrast to random variable theory, matrix algebra is assumed familiar, with only a brief appendix giving important results. Instead of classical cluster-analysis material, one finds a chapter on normal mixtures, and principal component analysis is presented in a novel way, via self-consistent approximation.
This book is a detailed work of scholarship, and anyone who reads through all the material should obtain an enviable coherent view of the subject. How-ever, many intended readers will find the theory hard going. Teachers will be greatly stimulated by the book, which is packed with exercises, but without solutions. It is a shame the index spans only two and a half pages.

Contents:
1. Introduction
2. Assumptions
3. EM and data augmentation
4. Inference by data augmentation
5. Methods for normal data
6. More on the normal model
7. Methods for categorical data
8. Loglinear models
9. Methods for mixed data
10. Further topics

Readership: Statisticians, biostatisticians, practitioners of sample surveys, graduate students and other methodologically-oriented researchers who have incomplete data

The aim of this book is to raise awareness and facilitate practical use of the great advances in methods for handling incomplete data which have been made in the last two or three decades. The book describes iterative algorithms for generating multiple imputations for missing values based on three multivariate models: the multivariate normal distribution for continuous data, log-linear models for categorical data, and the location model for mixed data types. In all cases the missing data are assumed to be missing at random. The author is aware of the limitations that these impose, but felt that a thorough treatment of the narrower area was appropriate.
The first four chapters provide a general outline of issues of incomplete data, covering topics such as ignorable and non-ignorable missingness, the EM algorithm, the MCMC approach, and inference based on augmented samples. The remainder of the book covers the three specific models already mentioned. The chapters concerned with specific models include numerical examples from real problems. Appendix C describes how to obtain (free of charge) S and Fortran-77 implementations of the algorithms.
The book is a pleasure to read and I recommend it to anyone who is beginning work on issues of incomplete data or who has problems involving incomplete data and wishes to resolve them in ways which are less susceptible to incorrect results than the approaches generally available in commercial software.

Contents:
1. Generalized linear modelling
2. Discrete data
3. Fitting and comparing probability distributions
4. Growth curves
5. Time series
6. Survival data
7. Event histories
8. Spatial data
9. Normal models
10. Dynamic models
APPENDIX A: Inference
APPENDIX B: Diagnostics

This book is compendium of models for data. Modelling is viewed as a search for a simplifying systematic pattern in empirical data containing random variation. The systematic pattern is usually of a regression type and the random part follows some specified probability distribution. With this abstract idea, the author produces a book of great practical importance.
A glance at the chapter headings shows the breadth of coverage, each topic being the subject of many books and articles in its own right. In each chapter a succinct review of the features of the data that motivate the models under discussion is given, followed by worked examples of the types of models that could be fitted by maximum likelihood and selected by the Akaike Information Criterion, (AIC). The implications and limitations of each model fitted are discussed, with many insightful comments. The discussion sometimes leads the reader to another approach in a later chapter. Further sets of data are given in the exercises that accompany each chapter.
The aim of this book is not to teach the reader the techniques of model fitting. There is no detailed technical information on these but the copious references to the literature would enable the reader to get started. For the more sophisticated, Chapter 10 in particular draws attention to many areas where further research is needed.
The value of this book is the way in which it will alert the statistical community and any gatherer of data to the scope and power of the modelling approach to the analysis of data. It is highly recommended.

Contents:
1. Introduction
2. The S-language
3. Graphical output
4. Programming in S
5. Distributions and data summaries
6. Linear statistical models
7. Generalized linear models
8. Robust statistics
9. Non-linear models
10. Random and mixed effects
11. Modern regression
12. Survival analysis
13. Multivariate analysis
14. Tree-based methods
15. Time series
16. Spatial statistics
17. Classification
APPENDIX A: Datasets and Software
APPENDIX B: Common S-PLUS Functions
APPENDIX C: Using S-PLUS Libraries

Readership: Aspiring, causal, and serious users of S-PLUS. Students and teachers of data analysis and statistics, statisticians

The first edition has lived up to this re-viewer's expectations [Short Book Reviews, Vol. 15, p.24], proving to be a reliable source of solutions to a variety of problems in teaching, consulting and research. The second edition improves on the first in several respects: almost one hundred pages of additional material, including chapters on random and mixed effects and classification; updates on new features of S-PLUS such as Trellis Graphics; and numerous improvements in the writing due to clarified exposition as well as new explanations, illustrations and practical suggestions. In place of a floppy disk with sets of data and supplementary software and an appendix containing answers to exercises, the authors list several World Wide Web sites where these features can be accessed along with on-line help for the software and on-line complements to the text.

Contents:
1. Introduction
2. Natural exponential families of Levy processes
3. Definitions and examples
4. First properties
5. Random time transformations
6. Exponential families of Markov processes
7. The envelope families
8. Likelihood theory
9. Linear stochastic differential equations with time delay
10. Sequential methods
11. The semi-martingale approach
12. Alternative deffinitions
APPENDIX A: A Toolbox from Stochastic Calculus
APPENDIX B: Miscellaneous Results
APPENDIX C: References
APPENDIX D: Notation

This book introduces students and researchers in statistics to inference for stochastic processes using the most tractable examples, exponential families. Much of the book is devoted to the study of these families of processes. Inference is approached via likelihoods, the sequential maximum likelihood estimator and sequential probability ratio test being highlighted. Exponential families of semimartingales are described in terms of their local characteristics. Each chapter is complemented with illuminating exercises and interesting historical notes. Appendices supply needed stochastic calculus and other stochastic process fundamentals. The book is a substantial addition to the as yet small book literature in this direction.

Contents:
1. The main single-server model
2. The main multiserver model
3. Advanced single-server models
4. Advanced multiserver models
5. Bibliographical remarks

Readership: Researchers, lecturers and graduate students in queueing theory and in areas of application of queueing theory

This book is a systematic presentation of the theory of retrial queues. It gives a clear and careful unified treatment of results collected from many different sources, and is a useful and readable introduction to the area. Essential prerequisites include probability theory and Markov chains; some basic queueing theory would be an advantage. Chapter 1 gives detailed results for the main single-server model with Poisson arrivals and exponentially distributed inter-repeat times, and in Chapter 3 these results are extended to systems with priority subscribers, impatient customers, etc. Chapters 2 and 4 have a different flavour from Chapters 1 and 3 and concentrate on the multiserver case with exponential service times. In these chapters, the authors include discussion of approximations and computational considerations, and there are several Pascal programs. An attractive feature of the book is Chapter 5, which is an extensive overview of the literature on retrial queues.

Contents:
1. Introduction and coverage
2. Probabilistic results in sequential estimation
3. Some basic concepts for fixed-sample estimation
4. General aspects of sequential estimation
5. Sequential Bayesian estimation
6. Multistage estimation
7. Parametric sequential point estimation
8. Parametric sequential confidence estimation
9. Nonparametric sequential point estimation
10. Nonparametric sequential confidence estimation
11. Estimation following sequential tests
12. Time-sequential estimation problems
13. Sequential estimation in reliability models
14. Sequential estimation of the size of a finite population
15. Stochastic approximation

Research on sequential analysis has led to the publication of many books and papers. Sequential estimation has received less attention. This volume tries to "present the theory of sequential estimation in a systematic and unified fashion at the graduate level", taking into account developments in the last fifteen years. It assumes familiarity with probability theory and statistical inference, though certain background results are reviewed. Chapters 3 and 4 contrast the fixed and sequential sampling approaches. This is followed by consideration of the Bayesian viewpoint. Two-stage/multistage estimation and accelerated sampling techniques (Chapter 6) lead naturally to the study of purely sequential point and interval estimation in parametric and nonparametric setups (Chapters 7-10).
The book then turns to more specialized topics. Time-sequential estimation (as in medical studies) is distinguished from sequential estimation by lack of an independent and identically distributed structure, stochastic cost-functions, possibly non-random sample sizes and possibly non-integer stopping times. Other topics are reliability models, such as bundle strength of filaments and system repair and availability models, and the sequential estimation of finite population sizes by capture-recapture sampling and sequential tagging. The final chapter examines sequential estimation in stochastic approximation, for example when successive dose levels are stochastic. There are many useful references. I had considerable problems with the name index.
This valuable seminal work crams a lot of material, including some not previously treated in depth in book form, into a medium-sized volume. The conciseness of the exposition is hampered by the unfriendly typeface and overcrowded pages.

Contents:
1. Introduction
2. Graphs and hypergraphs
3. Conditional independence and Markov properties
4. Contingency tables
5. Multivariate normal models
6. Models for mixed data
7. Further topics (Expert systems, model selection, modelling complexity, missing data)
APPENDIX A: Pre-requisites (Statistical Concepts)
APPENDIX B: Linear Algebra and Random Vectors
APPENDIX C: The Multivariate Normal Distribution
APPENDIX D: Exponential Models

Readership: Researchers and students in mathematical statistics, researchers in artificial intelligence and statistical physics, mathematicians, computer scientists

This book is an excellent reference both for the theory of graphical models and for the use of graphical models in statistics, although the book is primarily mathematical in nature, rather than being a manual for how to use graphical models. The author com-bines the state of knowledge in the field with many of his own extensive contributions to the theory of graphical models in statistics. The essential theory of graphical models is concisely outlined in Chapters 2 and 3, and then the statistical application of this theory in contingency tables is motivated by examples at the beginning of Chapter 4, followed by a comprehensive discussion of the mathematical theory of graphical models for contingency tables. The author carefully explains each proof, and provides asymptotic results where appropriate. Proofs are given in detail. Chapters 5 and 6 suffer a bit from lack of motivation, as to why a researcher might be interested in the methods and theory therein, but since this book is primarily intended as a mathematical reference, this is not a huge stumbling block. Chapter 7 peaks the reader's interests in other areas of graphical modelling, with a number of examples. The appendices do a nice job of summarizing some of the many mathematical and statistical concepts that are needed to work one's way through the material. Overall, this is an excellent reference book for those who are interested in studying the mathematical theory for graphical models.

Contents:
1. Some facts from regression analysis
2. Convex design theory
3. Numerical techniques
4. Optimal design under constraints
5. Special cases and applications

These authors are distinguished researchers in the area of optimal experimental designs, with most recent joint work in the area of spatial observation. This latter work shows its influence in this monograph in versions of optimal design duality threorems, and also algorithms, appearing in a more general form with regard to descriptions of designs measures and specifications of correlation functions. This is really a book for specialists but is a surprisingly fluid read for non-specialists compared to the sometimes very techni-cal work which appears in the recent literature. The monograph demonstrates the strong US-European axis in optimal design and is very much in the spirit of the MODA conferences, within which both authors have played pivotal roles.

Contents:
PART I : Efficient Inference for Planned Experiments
1. Planned experiments
2. Efficiency concepts for outlier-free observations
PART II : Robust Inference for Planned Experiments
3. Smoothness concepts of outlier robustness
4. Robustness measures: Bias and breakdown points
5. Asymptotic robustness for shrinking contamination
6. Robustness of tests
PART III: High Robustness and High Efficiency
7. High robustness and high efficiency of estimation
8. High robustness and high efficiency of tests
9. High breakdown point and high efficiency

Readership: Mathematical statisticians with interests in robust statistical techniques and experimental design

To my knowledge this is the first monograph where a quite successful attempt is made to marry two rather mature subjects of statistics: robust data analysis and experimental design. A very substantial amount of published results, including the author's, were accumulated and presented in a general setting. The book could attract more readers if interplay between two optimization problems (robust estimation and optimal design) were discussed and commented on in a less formalistic way and appealing more to the reader's intuition. The notation is menacingly complicated and the six-page list of symbols really helps one to move through the text which is very densely populated with formulae.

Contents:
1. Overview and preliminaries
2. First-order asymptotics
3. Second-order asymptotics
4. Extensions of results from the one-parameter case
5. Exponential family regression and diagnostics
6. Curvature in exponential family regression
7. Information-metric Riemannian geometry
8. Statistical manifolds
9. Divergence functions
10. Recent developments

This book gives self-contained foundations of differential-geometrical methods in statistical inference. It combines the geometrical methods with a variety of asymptotic analysis, which offers a unified idea on statistical inference from a geometric point of view. The methodology expanded in the book is sufficiently accessible for statisticians who are not so familiar with differential geometric tools. Statistically important examples have been discussed along with the progress of geometric considerations. Finally, recent developments with infinite-dimensional manifold theory are discussed.

Contents:
PART I : Differential Calculus on Gaussian Probability Spaces
1. Gaussian probability spaces
2. Gross-Stroock Sobolev spaces over a Gaussian probability space
3. Smoothness of laws
PART II : Quasi-sure Analysis
4. Foundations of quasi-sure analysis: Hierarchy of capacities and precise Gaussian probability spaces
5. Differential geometry on a precise Gaussian probability space
PART III: Stochastic Integrals
6. White noise stochastic integrals as divergences
7. Itô's theory of stochastic integration
PART IV : Stochastic Differential Equations
8. From ordinary differential equations to stochastic flow: The transfer principle
9. Elliptic estimates through stochastic analysis
PART V : Stochastic Analysis on Wiener Spaces
10. Stochastic analysis on Wiener spaces
11. Path spaces and their tangent spaces

Readership: Postgraduates and researchers in probability theory and analysis (partial differential equations to differential geometry)

This book spans a huge area ranging over stochastic processes, Gaussian measures of infinite-dimensional spaces, stochastic flows, potential theory and capacities, partial differential equations, and infinite dimensional differential geometry. It is written by an author who, starting with his solution of the stochastic version of Hörmander's hypoellipticity problem, is having a great influence on the interaction and confluence of big areas of lively mathematics which are still in fast development. The experience and knowledge the author gives may not be for all possible readers but certainly for those who know at least elements of the theory as a `big sky' experience. The book is written in a style which should motivate any "experienced worker" to have it on his bookshelf. I cite from the preface: `Our geometric point of view has obliged us to pay great attention to the foundations. On the other hand our notation, which follows the usual conventions, will allow any experienced worker to look directly at any section of the book, without spending time on the foundational sections.' The author has managed to keep up with this big challenge: from any point in the book one can get a very clear and often unexpected view into various areas of beautiful, yet demanding, mathematics. This made the book fun to read for me, and I can recommend it warmly.

Title	THE EM ALGORITHM AND EXTENSIONS. G.J. McLachlan and
Author	T. Krishnan.
Publisher	New York: Wiley, 1997, pp. xvii + 274, £50.00.

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

Title	PROBABILITY, STOCHASTIC PROCESSES, AND QUEUEING THEORY.
Author	R. Nelson.
Publisher	New York: Springer-Verlag, 1995, pp. xxviii + 583, US$34.95.

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

Title	SOME ASPECTS OF BROWNIAN MOTION, PART II: SOME RECENT MARTINGALE PROBLEMS.
Author	M. Yor.
Publisher	Basel: Birkhäuser, 1997, pp. xii + 144. DM.38.00/ÖS.278.00/Sw.fr.32.00.

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

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

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

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

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

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

Reviewer:
Institute	Carleton University
Place	Ottawa, Canada
Name	P. Calamai

Reviewer:
Institute	Queen's University
Place	Kingston, Canada
Name	A.M. Herzberg

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

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

Reviewer:
Institute	University of Oxford
Place	Oxford, U.K.
Name	P.J. Donnelly

Reviewer:
Institute	Harvard School of Public Health
Place	Boston, U.S.A.
Name	S.W. Lagakos

Reviewer:
Institute	London School of Hygiene, and Tropical Medicine
Place	London, U.K.
Name	K. Tomlin

Reviewer:
Institute	McMaster University
Place	Hamilton, Canada
Name	S.R. Esterby

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

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

Reviewer:
Institute	Dalhousie University
Place	Halifax, Canada
Name	C.J. Field

Reviewer:
Institute	University of Reading
Place	Reading, U.K.
Name	D.J. Balding

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

Reviewer:
Institute	Oak Ridge National Laboratory
Place	Oak Ridge, U.S.A.
Name	G. Ostrouchov

Reviewer:
Institute	University of Regina
Place	Regina, Canada
Name	R.J. Tomkins

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

Reviewer:
Institute	Erasmus University
Place	Rotterdam, The Netherlands
Name	L. de Haan

Reviewer:
Institute	University of Groningen
Place	Groningen, The Netherlands
Name	T. Mikosch

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

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

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

Reviewer:
Institute	__________
Place	Birmingham, U.K.
Name	D.M.G. Wishart

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

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