scientific article; zbMATH DE number 3846557

From MaRDI portal
Publication:3316288





The authors extend previous results on exact approximation orders in the conditional central limit theorem [Z. Wahrscheinlichkeitstheor. Verw. Geb. 66, 227-244 (1984; Zbl 0525.60029)] where the conditional distribution of \(S_ n=n^{-1/2}(X_ 1+...+X_ n)\) given a measurable set A was investigated. In this paper they consider i.i.d. random vectors \(X_ n\) taking values in \({\mathbb{R}}^ k\) with identity covariance matrix, mean zero and s-th absolute moment \(3\leq s<\infty.\) For classes H of conditioning functions with uniformly bounded \(L_ r\) norm where \(r=\infty\) if \(s=3\), and \(r>1+1/(s-3)\) otherwise, they prove for a class \({\mathcal F}\) of functions \(f:{\mathbb{R}}^ k\to [-1,1]\) such that the unconditional Berry-Esséen theorem holds uniformly, that \(\sup_{\phi \in H}dist(\phi,\sigma(X_ 1,...,X_ n))=O(n^{- \alpha}\log n^{\beta})\) implies \[ \sup_{f\in {\mathcal F},\phi \in H}| Ef(S_ n)\phi -Ef(s)E\phi | =O(n^{-1/2})\quad if\quad \alpha =1/2,\quad \beta<-3/2, \] \[ =O(n^{-1/2}\log \log n)\quad if\quad \alpha =1/2,\quad \beta =-3/2, \] \[ =O(n^{-1/2}(\log n)^{\beta +3/2})\quad if\quad \alpha =1/2,\quad \beta>-3/2, \] \[ =O(n^{-\alpha}(\log n)^{\beta +\alpha})\quad if\quad 0<\alpha<1/2. \] The authors give examples showing that these rates are optimal.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3316288)