点击次数：

论文类型：期刊论文

发表时间：2014-05-01

发表刊物：STATISTICS & PROBABILITY LETTERS

收录刊物：SCIE、Scopus

卷号：88

期号：1

页面范围：133-140

ISSN号：0167-7152

关键字：Gordon's inequality; Slepian's inequality; Comparison inequality

摘要：Let {xi(i,j)} and {eta(iota,j)}(1 <= i <= n, 1 <= j <= m) be standard Gaussian random variables. Gordon's inequality says that if E(xi(i,j)xi(i,k)) >= E(eta(i,j)eta(i,k)) for 1 <= i <= n, 1 <= j, k <= m, and E(xi(i,j)xi(l,k)) <= E(eta(i,j)eta(l,k)) for 1 <= i not equal l <= n, 1 <= j, k <= m, the lower bound P(boolean OR(n)(i=1) boolean AND(m)(j=1) {xi(i,j) <= lambda(i,j)})/P/(boolean OR(n)(i=1) boolean AND(m)(j=1)) {eta(i,j) <= lambda(i,j)}) is at least 1. In this paper, two refinements of upper bound for Gordon's inequality are given. (C) 2014 Elsevier B.V. All rights reserved.