scientific article; zbMATH DE number 4013651

From MaRDI portal
Publication:3025968





Let \(S_ n=X_ 1+...+X_ n\) be the sum of independent identically distributed random variables with E \(X_ i=0\), Var \(X_ i=1\). Necessary and sufficient conditions are indicated in order that \[ (1)\quad P(S_ n\geq x\sqrt{n})\leq e^{-\delta x} \] for \(0\leq x\leq \Lambda (n)\) and \(n\geq n_ 0\) where \(\delta >0\), \(\Lambda\) (n) is a function such that \(\Lambda\) (n)\(\to \infty\) as \(n\to \infty.\) Theorem 1. The inequality (1) with \(\Lambda (n)=\sqrt{n}/\Psi (n)\) where \(\Psi\) (n) is any function such that \(\Psi\) (n)\(\to \infty\) as \(n\to \infty\), holds iff E exp\(\{\) hX\({}_ 1\}<\infty\), \(0\leq h<H\), for some \(H>0.\) Theorem 2. Fix a number \(\alpha\), \(<\alpha <1\). The inequality (1) with \(\Lambda (n)=bn^{\alpha -}\), \(b>0\), is valid iff \(P(X_ 1>x)=O(\exp \{-cx^{2-1/\alpha}\})\), \(x\to \infty\), for some \(c>0\).




Cited in
(33)








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 Q3025968)