The weak convergence for functions of negatively associated random variables (Q5947228)

Consider a sequence of stationary, negatively associated random variables, and set \(S_j(1)= X_{j+k}+\cdots+ X_{j+k}\), \(S_n= X_1+\cdots+ X_n\), and \(Y_{k,j}= f(\{S_j(k)- k\mu\}/\sqrt k)\), with \(\mu\) denoting the mean of \(X_1\), and \(f\) a real function satisfying some conditions. The author's main results provide an asymptotic Wiener process approximation for the partial sum process of the \(\{Y_{k,j}: j=1,\dots, n\}\), both in the case of \(k\) being a fixed positive integer as well as for \(k\to \infty\), but \(k/n\to 0\) as \(n\to\infty\). A corresponding result for positively associated random variables is also presented. As a consequence of the main results, the author is able to derive asymptotic normality of some estimators of the asymptotic variance of \(S_n\), which have earlier been discussed by \textit{M. Peligrad} and \textit{Q.-M. Shao} [ibid. 52, No. 1, 140-157 (1995; Zbl 0816.62027)] and \textit{M. Peligrad} and \textit{R. Suresh} [Stochastic Processes Appl. 56, No. 2, 307-319 (1995; Zbl 0817.62019)] in case of \(\rho\)-mixing variables, and by \textit{Zhang} and \textit{Shi} (1998) for negatively associated variables.