Comportement asymptotique du temps d'occupation du processus des sommes partielles. (Asymptotical behavior of the occupation time of partial sums) (Q1209196)

Consider a sequence \((e_ t, t\in\mathbb{N})\) of i.i.d. real-valued variables and the process of its partial sums, \(X_ t=\sum^ t_{k=1}e_ k\), so \(X\) is a random walk. The paper is concerned with strong approximation results for the occupation times of \(X\) by those of a Wiener process. An extension of a theorem of \textit{A. N. Borodin} [Probab. Theory Relat. Fields 72, 231-250, 251-277 (1986; Zbl 0572.60078 and Zbl 0572.60079)] is proved. Namely, suppose that \(Ee_ 1=0\), \(Ee^ 2_ 1=1\) and \(E| e_ 1|^ p<\infty\) for some \(p>2\). If \(\lim_{t\to\infty}t^ \beta E(\exp(ite_ 1))=0\) for some \(\beta>0\), then there exists a Brownian motion \(W\) and a random walk \(X'\) distributed as \(X\) such that, for all \(a\in\mathbb{R}\) and any sequence \((\delta_ n)\) with \(\delta_ n\geq n^{-1/6-2/3p}\), \[ \sum^ n_{i=1}1_{[a,a+\delta_ n]}(X_ i')- \int^ n_ 01_{[a,a+\delta_ n]}(W_ s)ds= o(\delta_ n^{1/3+1/3p+\varepsilon})\text{ a.s. } \] for all \(\varepsilon>0\). The proof relies on an extension of Gnedenko's local limit theorem. Another result in the same vein is obtained for the process of the partial sums of a linear stationary process \((\xi_ t,t\in\mathbb{N})\).