An invariance principle for negatively associated random variables (Q1370170)

For a strictly stationary sequence of negatively dependent r.v.'s \(X_j\), \(j\in\mathbb{N}\), with finite variance the assumption that \(L(n)=\text{var }X_1+2\sum_{j=1}^n\text{cov}(X_1,X_j)\) is a slowly varying function (instead of requiring \(A=\text{var }X_1+2\sum_{j=1}^\infty\text{cov}(X_1,X_j)\)) permits to extend the invariance principle established by \textit{C. Su} et al. [Sci. China, Ser. A 40, No. 3, 347 (1997)]. The weak convergence of the partial sum processes (involving the normalization \((nL(n))^{1/2}\) instead of \((nA)^{1/2}\)) to the Wiener process in the space \(C[0,1]\), \(T>0\), is shown to be equivalent to a certain restriction on tails of distributions of \(S_n/(nL(n))^{1/2}\) for the specified subsequence \(\{n_1\}\). Reviewer's remark. In the definition of negative association one has to consider the covariance of functions (non-decreasing in each variable) for disjoint sets of random variables.