Entropy inequalities and the Central Limit Theorem

被引:15
|
作者
Johnson, O [1 ]
机构
[1] Univ Cambridge, Stat Lab, Cambridge CB2 1SB, England
基金
英国工程与自然科学研究理事会;
关键词
normal convergence; entropy; Fisher information;
D O I
10.1016/S0304-4149(00)00006-5
中图分类号
O21 [概率论与数理统计]; C8 [统计学];
学科分类号
020208 ; 070103 ; 0714 ;
摘要
Motivated by Barren (1986, Ann. Probab. 14, 336-342), Brown (1982, Statistics and Probability: Essays in Honour of C.R. Rao, pp. 141-148) and Carlen and Soffer (1991, Comm. Math. Phys. 140, 339-371), we prove a version of the Lindeberg-Feller Theorem, showing normal convergence of the normalised sum of independent, not necessarily identically distributed random variables, under standard conditions. We give a sufficient condition for convergence in the relative entropy sense of Kullback-Leibler, which is strictly stronger than L-1. In the IID case we recover the main result of Barren [1] (c) 2000 Elsevier Science B.V. All rights reserved.
引用
收藏
页码:291 / 304
页数:14
相关论文
共 50 条