Asymptotic recurrence and waiting times for stationary processes (Q1266775): Difference between revisions

Let \(\{x_n\}\) be a realization of a discrete-valued stationary ergodic process \(\{X_n\}\) with distribution \(P\). Define \(x_i^j=(x_i,\dots,x_j)\) and \(R_n=\inf\{k\geq 1: x_{-k+1}^{-k+n}=x_1^n\}\). If \(\{x_n\}\) and \(\{y_n\}\) are two independent realizations of \(\{X_n\}\), define \(W_n= \inf\{k\geq 1:y_k^{k+n-1}=x_1^n\}\). The variables \(R_n\) and \(W_n\) are called recurrence time and waiting time, respectively. The author proves a strong approximation theorem between \(\log R_n\) and \(-\log P(X_1^n)\) and deduces an almost sure invariance principle for \(\log R_n\). Similar results are proved also for \(W_n\).