Invariance principles and Gaussian approximation for strictly stationary processes

From MaRDI portal





Let \((X,{\mathcal F},T,\mu)\) be an aperiodic dynamical system. It is shown that one can find a square integrable function \(f:X\to \mathbb{R}\) such that the normalized partial sum process \(S_n(f)= (f+\cdots+ fT^{n-1})/ \sqrt n\) can be approximated by the normalized partial sums \(W_n\) of an i.i.d. normally distributed sequence (defined on a suitable common probability space if necessary). (1) The distance in \(L_2\)-metric is of rate \(O(\log \log\log\log n/\log n)^{1/2}\) where the normal distribution is centered and has variance \(2(\log\log 3-\log\log 2)\). (2) The distance between the laws tends to zero (weak invariance principle), where the normal distribution is the standard one. (3) The Euclidean distance is of order \(O(\log \log n)^{1/2}\), where the normal distribution is the standard one. This result extends those of \textit{R. Burton} and the reviewer [Trans. Am. Math. Soc. 302, 715-726 (1987; Zbl 0628.60030)] and \textit{M. Lacey} [Trans. Am. Math. Soc. 328, 767-778 (1991; Zbl 0741.60026)].




Cited in
(23)








This page was built for publication: Invariance principles and Gaussian approximation for strictly stationary processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4243648)