###
Announcement

## Multivariate Statistical Analysis. A High-Dimensional
Approach

###
by V.Serdobolskii

#### published by "Kluwer Science Publishers" in October of 2000

ISBN 0-7923-6643-3

#### Mailing address:

Kluwer Academic Publishers, Sales Departement

PO BOX 17, 3300 AA, Dordrecht, The Netherlands

#### Site to order:

#### www.wkap.nl/boordfrm?0-7923-6643-3+1

In contrast to other books on multivariate analysis
retelling
classical theory of the multivariate statistical analysis based on the
idea of a consistency, the author develops an essentially new
theoretical-decision approach to the multivariate analysis that
leads to always stable
approximately unimprovable distribution free methods.

This book may be of
interest for specialists in mathematical statistics in universities
and institutes with mathematical departments, departments of statistics,
research institutes and
computer centers creating and improving packages of applied statistical
programs. The book can serve as a text-book for students and
post-graduates studying mathematics and statistics.

The book "Multivariate Statistical Analysis. A High-Dimensional Approach"
presents a new branch of mathematical
statistics solving problems of construction of approximately
unimprovable versions of the multivariate analysis methods such
as multiparametric estimation, discriminant and regression analysis.
In contrast to the traditional consistent Fisher method of statistics,
the essentially multivariate technique is developed based on the decision
function approach by A. Wald. For
large dimension of observations comparable in magnitude with sample size,
stable approximately unimprovable in some wide classes procedures
depending on an arbitrary function are offered. A remarkable fact
is established: for large-dimensional problems, under some weak
restrictions on the variable dependence, the standard quality functions
of regularized multivariate procedures prove to be independent of
distributions. This opens, for the first time in the history of statistics,
a possibility to construct approximately unimprovable
procedures free from distributions.