Uniform CLT for empirical process (Q2485830)

The aim of this paper is to provide sufficient conditions such that stochastic equicontinuity holds for an empirical process indexed by a set of functions \({\mathcal F}\) and being based on (strictly) stationary dependent random variables. Stochastic equicontinuity is the basic ingredient for weak convergence of such processes on \(\ell_\infty({\mathcal F})\,.\) Two basic conditions are needed here, a Rosenthal type moment inequality of order \(p \geq 2\) for the empirical process and an integrability condition on the bracketing number of \({\mathcal F}\) using a norm related to the first condition. As applications, results for empirical processes based on \(\alpha\)-mixing, \(\rho\)-mixing and Gaussian sequences of random variables are given, thereby improving some results from the literature.