Multiple decorrelation and rate of convergence in multidimensional limit theorems for the Prokhorov metric. (Q1889789): Difference between revisions

Let \((\Omega,{\mathcal F},\nu,T)\) be a probabilistic dynamical system, i.e. \((\Omega,{\mathcal F},\nu)\) is a probability space and \(T:\Omega\rightarrow\Omega\) a \(\nu\)-preserving transformation. Suppose that \(\Omega\) is endowed with a metric and that \({\mathcal F}\) is the associated Borel-\(\sigma\)-field. For any integer \(d\geq1\) and measurable function \(f:\Omega\rightarrow R^d\) define \(S_n=\sum_{k=0}^{n-1}f\circ T^k\) for every integer \(n\geq1\). If \((\Omega,{\mathcal F},\nu,T)\) satisfies a multiple decorrelation condition whose formulation is technically too involved to be repeated here, and if \(f\) is a \(\nu\)-centered, bounded Hölder continuous function, then it is shown that the limit \(D(f)\) of the covariance matrices of \(S_n/\sqrt{n}\) exists and that the Prokhorov distance between the distribution of \(S_n/\sqrt{n}\) and the \(d\)-dimensional centered normal distribution with covariance matrix \(D(f)\) is of the order \(O(n^{-1/2+\alpha})\) for all \(\alpha>0\). The proof uses characteristic function techniques and a smoothing inequality for the Prokhorov distance of \(d\)-dimensional distributions due to \textit{V. V. Yurinskij} [Theory Probab. Appl. 20, 1--10 (1975); translation from Teor. Veroyatn. Primen. 20, 3--12 (1975; Zbl 0351.60007)]. In the second part it is shown that a variant of the method can be used to derive a bound in a central limit theorem for the error term between the solution of a differential equation perturbed by a transformation and the solution of the associated averaged differential equation.