The multivariate normal distribution (Q1188605)

This book gives an excellent survey of the properties of the multivariate normal (MVN) distribution. It is not the purpose of the author to complement or to competewith books on multivariate analysis: ``... we shall not make an attempt to include inference problems and data analysis methods in this book. Instead, we shall confine our attention to the distributional aspects of the MVN theory.'' The book is divided into 9 Chapters, each concluding with a useful list of problems. After an introductory Chapter 1 the fundamental properties of the bivariate and multivariate normal distributions are presented in Chapters 2 and 3. This includes characterizations as well as principal components and correlation analysis and the basic sampling distributions. In Chapter 4, properties related to the shape of the MVN density are investigated, as for instance its relations to elliptically contoured distributions, generalizations of total positivity, Schur concavity, and arrangement increasing properties. Chapter 5 is devoted to positive dependence and exchangeability of the components of a normal random vector. The order statistics of these components are investigated in Chapter 6 with emphasise on the influence of the original dependence structure to the ordered components. Chapter 7 presents probability and moment inequalities for the MVN distribution. Chapter 8 is entitled ``Statistical Computing'' and includes methods for generating MVO variates and methods for approximating MVN probability integrals. Chapter 9 contains the distributional properties of the multivariate t distribution. Using the methods of Chapter 8 tables of the equicoordinate one-sided and two-sided percentage points and probability integrals have been computed and can be found in the Appendix. In the reviewer's opinion, this book stands in the tradition of the books of \textit{N. L. Johnson}, and \textit{S/ Kotz} Distributions in Statistics: Continuous multivariate distributions (1972; Zbl 0248.62021), though on the MVN topic it goes beyond as most results are proved and numerous examples of various kinds are given. The book is well organized and very clearly written. Since it only requires a basic knowledge of matrix algebra and mathematical statistics it can be used as a main reference in a course on multivariate analysis for students of mathematics, statistics and related applied areas. Furthermor, the book also contains about 200 references and various hints for those readers who may have a deeper interest in a specific topic. One topic the reviewer sadly missed is a treatment of the multivariate central limit problem; it could have made a further interesting chapter of the book. In summary, the book under review provides a well-written presentation of annotated distributional properties on the MVN distribution that should be useful as a reference text for students as well as for researchers.