No label defined (Q3759608)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 4009421
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 4009421

      Statements

      0 references
      1987
      0 references
      FKG inequality
      0 references
      majorization
      0 references
      permutation orbit
      0 references
      Schur concave functions
      0 references
      applications of majorization to the problems of mathematical statistics
      0 references
      log concavity
      0 references
      convolution of symmetric unimodal functions
      0 references
      Gauss-Markov estimators
      0 references
      monotonicity properties
      0 references
      Marshall-Olkin convolution theorem
      0 references
      The material of this book is based on a set of lectures given by the author in the University of Amsterdam. Three great topics are discussed in this book: majorization; the so-called FKG inequality (this name is due to the work of C. M. Fortuin, J. Ginibre and P. N. Kasteleyn, who gave sufficient conditions for (2), see below) and related topics. The unificating idea is undoubtedly that of majorization. It should be particularly mentioned that majorization was already the topic of a special monograph ''Inequalities: theory of majorization and its applications'' by \textit{A. W. Marshall} and \textit{I. Olkin}, (1979; Zbl 0437.26007). This book contains the most full material on majorization and its applications. In particular, a great section is devoted to the problems of mathematical statistics. The distinguishing features of this book are the geometrical approach (the author follows it all along the book) and great number of new examples of applications of the idea of majorization. It is written in an understandable way and in our opinion is the most suitable for the initial acquaintance with the subject. It will undoubtedly favour the propaganda of a promising method of majorization among specialists and students. Now we briefly shall chapter-wise dwell on the contents of the book. The first chapter is propaedeutical and contains the basic concepts of stochastic ordering for random variables later subjected to generalization for random vectors. The second chapter contains the description of the main characteristics of majorization in a finite-dimensional Euclidean space. As a basis of the definition of majorization the author uses a concept of the permutation orbit of the vector that is the set of vectors that are obtained with the help of the permutation of the coordinates of the initial vector. The proof of the fact that this definition of majorization called geometrical is equivalent to the traditional analytical one is given. The concept of majorization is closely connected with some special functions usually called Schur concave functions (the author calls them decreasing functions). In terms of these functions the concept of majorization enables to unificate the deduction of many known inequalities. Main properties of such functions and also an interesting convolution theorem due to Marshall and Olkin are given. This theorem states the conditions under which the convolution of Schur concave functions is Schur concave. The final part of the chapter contains possible generalizations of the definition of majorization and basic results for the special subsets of a Euclidean space. Some applications of majorization to the problems of mathematical statistics are given in the third chapter. In particular, these problems are connected with the characterization of multiparametric families of probability distributions. These families are characterized with the help of Schur concave functions possessing a special property. The author calls them decreasing reflection functions. This chapter contains rather many statements due to the author. The fourth chapter contains the description of the basic properties of log concave functions. The central result is Prekopa's theorem about the log concavity of a function that is represented as an integral of another log concave function. Some other theorems related with Prekopa's theorem are also contained in this chapter. Another topic of the chapter is the concept of unimodality. The presented results are grouping around the concept of A-unimodality (after Anderson, who introduced it) and Anderson's theorem about the convolution of symmetric unimodal functions. An analogy of majorization with the concept of invariance of subsets of a Euclidean space with respect to groups of transformations that are subgroups of the orthogonal group is drawn. The idea of majorization leads to a concept of a probability measure more concentrated about zero than another. This ordering of probability measures is attracted to the solution of some problems of mathematical statistics, in particular, to the study of concentration properties of Gauss-Markov estimators in certain linear models and the monotonicity properties of the power functions of some classical tests in MANOVA problems. Two possible ways of comparing random variables (vectors) are discussed in the fifth chapter. A random vector \(X=(X_ 1,...,X_ n)\) is said to be stochastically smaller than a random vector \(Y=(Y_ 1,...,Y_ n)\) if (1) Ef(X)\(\leq Ef(Y)\) for all coordinate-wise non-decreasing functions f(x), \(x=(x_ 1,...,x_ n)\in R_ n\). The other concept, called ''association'' by the author, is defined by the condition (2) \(cov\{f_ 1(X),f_ 2(X)\}\geq 0\) for all bounded coordinate-wise non-decreasing functions \(f_ 1\) and \(f_ 2\). Some classes of random vectors satisfying (1) or (2) are described in the chapter. Finally, Pitt's theorem about the conditions equivalent to the positiveness of all the elements of the covariance matrix of a normal random vector is proved. The sixth chapter contains some generalizations of majorization. The essence of these generalizations is the replacement of a group of permutation matrices by a more general subgroup of the orthogonal group. The author shows that many results listed above remain valid after such generalizations of the concept of majorization. In particular it regards the mentioned Marshall-Olkin convolution theorem.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references