scientific article; zbMATH DE number 3876409

From MaRDI portal
Publication:3341683





Let \(U_ 1,...,U_ n\) be independent r.v.'s which are uniformly distributed on the interval (0,1) and \(U_{1n}\leq...\leq U_{nn}\) be the corresponding order statistics. Consider the r.v. \(S_ n=\sum^{n}_{k=1}c_{kn}U_{kn}\), where \(c_{kn}\) are arbitrary constants. The Lévy concentration function of the r.v. \(S_ n\) is denoted by \(Q(S_ n;\lambda)=\sup_{x\in R}P\{S_ n\in [x,x+\lambda]\} (\lambda >0)\). It is proved that \[ Q(S_ n;\lambda)<C\lambda (1+1/n)(n+1)^{3/2}\Delta_ n^{-1/2}, \] where C is an absolute positive constant and \(\Delta_ n=\sum^{n}_{k=1}\sum^{n}_{m=k}(c_{kn}+...+c_{mn})^ 2.\) As corollaries of this result some sufficient conditions for the inequality \(Q(S_ n;\lambda)\leq C\lambda /\sqrt{n}\) are given.











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 Q3341683)