On weighted \(U\)-statistics for stationary processes. (Q1879839): Difference between revisions

A weighted \(U\)-statistic based on a random sample \(X_1, .., X_n\) has the form \[ U_n= \sum_{1\leq i, j \leq n} w_{i-j} K(X_i, X_j), \] where \(K\) is a fixed symmetric function and \(w_i\) are symmetric weights. This paper develops two general central limit theorems for statistics of this form to deal with cases where the weights are absolutely summable and cases where the weights are not summable. The theorems are applied to nonlinear time series that are geometric moment-contracting and to both short and long memory linear time series. The results for short memory processes have some overlap with results by \textit{S. Borovkova, R. Burton} and \textit{H. Dehling} [Trans. Am. Math. Soc. 353, 4261-4318 (2001; Zbl 0980.60031)], however the two sets of results have somewhat different ranges of applicability.