New techniques for empirical processes of dependent data (Q734659)

Let \((X_n)_{n\geq 0}\) be a stationary ergodic process of \(\mathbb{R}\)-valued random variables with marginal distribution function \(F(t)\). Define the empirical distribution function \((F_n(t))_{t\in\mathbb{R}}\) and the empirical process \((U_n(t))_{t\in\mathbb{R}}\). In this paper, the authors prove the empirical process invariance principle for the process \((X_n)\) under the following two conditions: 1. For any Lipschitz function \(f\), the CLT holds, i.e. \[ {1\over\sqrt{n}} \sum^n_{i=1} (f(X_i)- Ef(X_i))@>{\mathcal D}>> N(0,\sigma^2), \] where \[ \sigma^2= E(f(X_i)- Ef(X_i))^2+ 2\sum^\infty_{i=1} \text{Cov}(f(X_0), f(X_i)). \] 2. For \(f\) bounded Lipschitz with \(Ef(X_0)= 0\), \[ \begin{multlined} E\Biggl\{\sum^m_{i= 1} f(X_i)\Biggr\}^4\leq C\max\Biggl\{1, \sup_x|f(x)|\Biggr\} (n\| f(X_0)\|_1\log^\alpha(1+\| f\|)+\\ n^2\| f(X_0)\|^2_1\log^\beta(1+\| f\|)),\end{multlined} \] where \(C\) is some universal constant, \(\alpha\) and \(\beta\) are some nonnegative integers, and \[ \| f\|= \sup_x|f(x)|+ \sup_{x\neq y} {|f(x)- f(y)|\over |x-y|}. \] Some examples are also shown.