Characterizations of normality in translation classes by properties of Bayes estimators (Q1104656)

This paper generalizes some characterizations of the normal distribution due to \textit{A. M. Kagan} and \textit{Yu. N. Karpov} [J. Sov. Math. 3, 766- 776 (1975); translation from Zap. Nauchn. Semin. Leningr. Otd. Mat. Inst. Steklova 29, 62-73 (1972; Zbl 0337.62022)]. The characterizations under consideration concern the linearity of the Bayes estimator and the linearity of certain conditional expectations. Furthermore, the linearity of the best polynomial estimator of degree \(\leq r\) is derived as a characteristic property of a distribution family for which the first r moments coincide with the corresponding moments of a normal distribution. The original results of Kagan and Karpov were proved for independent variables, whereas the theorems given in the paper are valid in a certain linear model.