Weak convergence of weighted empirical \(U\)-statistics processes for dependent random variables (Q1319956)

Let \(\{X_ i\}\) be a strictly stationary sequence of real-valued random variables with distribution function \(F\). We assume that \(\{X_ i\}\) is absolutely regular, i.e., \[ \sup_ j \sup_{A \in \sigma (X_{j+m+1} \dots)} | P (A \mid \sigma(X_ 1,\dots,X_ j))-P(A) | \to 0, \] where \(\sigma (X_ a,\dots,X_ b)\) denotes the \(\sigma\)-algebra generated by \(X_ a,\dots,X_ b\). Let \(h:R^ k \to R\) be a symmetric Borel measurable function. Put \[ \begin{aligned} H_ n (t) &= {n \choose k}^{- 1} \sum_{i_ 1< \dots<i_ k} I\bigl(h(X_{i_ 1},\dots,X_{i_ k}) \leq t\bigr),\\ H(t) &= \int I\bigl(h(x_ 1, \dots,x_ k) \leq t\bigr) \prod^ k_{j=1} F(dx_ j),\quad t \in R,\end{aligned} \] and define the empirical \(U\)-statistic process \(\hat U_ n(t)=\sqrt n(H_ n(t)-H(t))\). The authors obtain the weak convergence of the weighted empirical processes of type \(U_ n(u)/r(u)\) \((\{U_ n\}\) being modified empirical \(U\)-statistical processes defined by \(\{U_ n\})\). The results in this paper extend those of \textit{F. H. Ruymgaart} and \textit{M. C. A. van Zuijlen} [J. Stat. Plann. Inference 32, No. 2, 259-269 (1992; Zbl 0754.62032)] for i.i.d. random variables.