数学物理学报 ›› 1997, Vol. 17 ›› Issue (4): 375-381.

• 论文 • 上一篇    下一篇

非负整值随机变量序列的一类强偏差定理

刘文   

  1. 河北工业大学数理系 天津 300130
  • 收稿日期:1995-07-10 修回日期:1995-11-19 出版日期:1997-08-26 发布日期:1997-08-26

A Kind of Strong Deviation Theorem for the Sequences of Nonnegative Integer-Valued Random Variable

Liu Wen   

  1. Department of Mathematics and Physics, Hebei University of Technology, Tianjin 300130
  • Received:1995-07-10 Revised:1995-11-19 Online:1997-08-26 Published:1997-08-26

摘要: 设{Xn,n ≥ 1}是在S={0,1,2,…}中取值的一列随机变量,其联合分布为fn(x1,…,xn),(p(0),p(1),…)是S上的一个分布,mk=1kp(k),pn(x1,…xn)=Πk=1np(xk),ψn=(1/nk=1n(Xk-m).该文研究对数似然比Ln(ω)=ln[pn(X1,…,Xn)/fn(X1,…,Xn)]与ψn(ω)之间的若干极限关系,得到了一类用不等式表示的强极限定理(称之为强偏差定理),其偏差界依赖于样本点.证明中结合区间刻分法,提出了将母函数的工具应用于强极限定理研究的一种途径.

关键词: 强偏差定理, 强极限定理, 似然比, 对数似然比, 母函数

Abstract: Let {Xn,n ≥ 1} be a sequence of random variables taking valnes in S={0,1,2,…} with the joint distribution fn(x1,…,xn),(p(0),p(1),…) a distribution on S,mk=1kp(k),pn(x1,…xn)=Πk=1np(xk), and ψn=(1/nk=1n(Xk-m). In this paper the limit relation between ψn(ω) and the log-likelihood ratio Ln(ω)=ln[pn(X1,…,Xn)/fn(X1,…,Xn)] is investigated and a kind of strong limit theorem represented by inequalities which we call the strong deviation theorem is obtained. In the proof an approach of applying the tool of generating function together with the method of splitting intervals to the investigation of the strong limit theorem is proposed.

Key words: Strong deviation theorem, Strong limit theorem, Likelihood ratio, Log-likeli-hood ratio, Generating function