Measurable supports, reducible spaces and the structure of the optimal \(\sigma\)-field in unbiased estimation (Q1068476)

This paper generalizes a result by \textit{P. N. Sapozhnikov} [Teor. Veroyatn. Primen. 23, 836-842 (1978; Zbl 0407.62016); English translation in Theory Probab. Appl. 23, 805-811 (1979)] relating the structure of the optimal \(\sigma\)-field in unbiased estimation to the reducibility of certain linear subspaces of estimators. For this purpose, the concept of a measurable support of (uncountably infinite) sets consisting of random variables is introduced, which could be of general interest. It turns out, that for a wide class of measure spaces, measurable supports of any subsets of measurable functions exist.