Identifiability of countable mixtures of discrete probability distributions using methods of infinite matrices

Abstract

Introduction. The subject of the mixtures of probability distributions has aroused a fresh interest in the fields of probability and statistics. Several authors, including Blischke, Rider, Robbins, Teicher, Patil and Seshadri, have discussed a variety of the associated probabilistic and statistical problems in varying details in the recent past. This paper attempts to solve by using the methods of infinite matrices some of the problems of the identifiability of countable mixtures of discrete probability distributions. It appears that here is still another area where Mathematics and Statistics can help one another grow. We are extremely thankful to Dr P. Vermes of Birkbeck College of the University of London for developing some theory of infinite matrices relevant to a couple of problems discussed in this paper. While most of the results discussed below are believed to be new, a few of them can be shown to be immediate consequences of a powerful result due to Teicher ((7)) obtained by using a different approach.