On the constant in the nonuniform version of the Berry-Esseen theorem

From MaRDI portal
Publication:2575997





Let \(X_1,X_2,\ldots,X_n\) be independent random variables with zero means such that \(W=\sum_{i=1}^nX_i\) has unit variance. Let \(F\) denote the distribution function of \(W\) and \(\Phi\) that of the standard normal distribution. \textit{L. H. Y. Chen} and \textit{Q.-M. Shao} [Probab. Theory Relat. Fields 120, 236--254 (2001; Zbl 0996.60029)] have shown that there exists an absolute constant \(C\) such that for all real \(x\) \[ | F(x)-\Phi(x)| \leq C\sum_{i=1}^n\Bigl\{{EX_i^2I(| X_i| >1+| x| ) \over (1+| x| )^2}+{E| X_i| ^3I(| X_i| \leq1+| x| )\over(1+| x| )^3}\Bigr\}\,. \] In the paper under review this result is made completely explicit by providing several numerical bounds on the constant \(C\), valid for different ranges of \(x\)-values. As in the paper by Chen and Shao, the approach is via Stein's method and a concentration inequality.











This page was built for publication: On the constant in the nonuniform version of the Berry-Esseen theorem

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2575997)