Simultaneous selection for homogeneous multinomial populations based on entropy function: An empirical Bayes approach

NCU Institutional Repository > 理學院 > 統計研究所 > 期刊論文 > Item 987654321/27802

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/27802

题名:	Simultaneous selection for homogeneous multinomial populations based on entropy function: An empirical Bayes approach
作者:	Gupta,SS;Leu,LY;Liang,TC
贡献者:	統計研究所
日期:	1996
上传时间:	2010-06-29 19:33:56 (UTC+8)
出版者:	中央大學
摘要:	Consider k independent m-cell multinomial populations pi(1),...,pi(k), where pi(i) has the associated cell probability vector (p) under tilde(i) = (p(il),...,p(im)), i = 1,...,k. The entropy theta(i) = theta((p) under tilde p(i)) = -Sigma(j=1)(m)p(ij) log p(ij) is used as a measure of homogeneity of the population pi(i). A Bayes selection rule relative to the Dirichlet prior distribution D-m(alpha(1),...,alpha(m)) is obtained for selecting all homogeneous populations compared with a standard theta(0). When the values of the parameters alpha(1),...,alpha(m) are unknown, an empirical Bayes selection rule is proposed and its corresponding asymptotic optimality is investigated. It is shown that the proposed empirical Bayes selection rule is asymptotically optimal relative to the class of Dirichlet prior distributions, and the associated regret risk converges to zero with a convergence rate of order O(exp(-tau k + ln k)) for some positive constant tau = tau(alpha(1),...,alpha(m)).
關聯:	JOURNAL OF STATISTICAL PLANNING AND INFERENCE
显示于类别:	[統計研究所] 期刊論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	394	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing