On the global central limit theorem for \(M\)-dependent random variables (Q1901202)

Let \(X_1, X_2, \ldots,X_n\) be real \(m\)-dependent random variables with \(EX_j = 0\) and \(E |X_j |^k < \infty\) for some \(k \geq 2\) and all \(j = 1,2, \ldots, n\). A number of experts in probability theory, in particular, \textit{R. V. Erickson} [Ann. Probab. 2, 522-529 (1974; Zbl 0301.60020)] and \textit{V. V. Shergin} [Teor. Veroyatn. Mat. Stat. 29, 122-128 (1983; Zbl 0577.60024)] proved some upper estimates of the quantity \(\int^\infty_{-\infty} |x |^l |P (\sum^n_{k = 1} X_k < xB_n) - \Phi (x) |dx\), where \(B^2_n = E (\sum^n_{k = 1} X_k)^2\) and \(\Phi\) is the standard normal distribution function. These estimates depend upon a non-negative number \(l,m\) and some other parameters. The author indicates the dependence of these estimates upon the parameter \(m\).