On conditions of existence of sums of functions of conditioned random elements (Q997727)

Let \((\Omega,\mathcal S,\mathbb P)\) and \((\mathcal F,\mathcal B)\) be probability and measurable spaces, respectively. Any \((\mathcal B,\mathcal S)\)-measurable mapping of a set \(\Omega\) into a set \(\mathcal F\) is called a random element. Let \(w_1,w_2,\dots,w_N\) be random elements given on a probability space \((\Omega_1,{\mathcal S}_1,\mathbb{P}_1\) and with values in the space \((\mathcal F,\mathcal B)\). Let us consider the sum: \[ \eta_N=\sum_{k=1}^Nf_k(w_k,w_{k+1},\ldots,w_{k+l}), \tag{1} \] where the summation in the index happens by a module \(N\), \(l\) is a positive integer, \(f_k\) are functions with values in a measurable space \(({\mathcal F}_0,{\mathcal B}_0)\), in which the operation of addition, continuous in some topology \({\mathcal U}_0\), is defined and \({\mathcal B}_0\) is a \(\sigma\)-algebra, caused by the topology \({\mathcal U}_0\). If causing random elements are dependent and spaces of values are locally compact Hausdorff abelian groups, then conditioned random variables represent a scheme of decomposable statistics and sums (1) are decomposable statistics. Such sums are useful in different contexts and the author notes that it is necessary to investigate more general sums (1), which can be studied similarly to decomposable statistics. In this direction, an analogue of the theorem of existence given in [\textit{E. M. Kudlaev}, Probabilistic methods in discrete mathematics. Proceedings of the third Petrozavodsk conference. Petrozavodsk (Russia), May 11--16, 1992. Moskva: TVP. Prog. Pure \& Appl. Discret. Math. 1, 271--286 (1993; Zbl 0813.60012)], is proved in two cases: causing elements are finitely dependent and Markov dependent. Moreover, ways of obtaining limit theorems for sums (1) are discussed.