Statistical problems of the Neyman type and decomposable statistics (Q1273366)

We consider estimators of distribution parameters, obtained by the moment method and the maximum likelihood method from a sample of specially dependent random variables; namely, we assume that the sample forms a scheme of decomposable statistics. This means the existence of independent random variables whose joint conditional distribution (under the condition that their sum is fixed at some ``point'') coincides with the joint distribution of the terms of that sample. The scheme of decomposable statistics (DSs) constitutes a rather wide class of dependent random variables, which appear both in various branches of mathematics and some other sciences. The DSs enjoy wide use in mathematical statistics in constructing goodness-of-fit and homogeneity tests or in survey sampling of finite populations. DSs surely arise while constructing Neyman-type tests based on a sufficient statistic for the nuisance parameter \(\vartheta\) for hypotheses on the informative parameter \(\xi\) (maybe, vector-valued). The maximum likelihood method due to Fisher for estimating distribution parameters is usually exploited in the classical situation, where the elements of the sample are independent and identically distributed by some law determined by a parameter. Only a few cases are known where the sample elements are assumed to be dependent. As far as the author knows, the problem posed in this paper is new, and the results obtained are of a rather general nature, as in the classical case.