完全随机缺失数据下配对试验的Bayes分析  

A Bayesian Method for Paired Design with Random Missing Data

作　　者：胡思贵[1,2] 赵明[2] 

机构地区：[1]贵阳医学院基础医学院,贵州贵阳550004 [2]贵州大学贵州省可靠性工程中心,贵州贵阳550025

出　　处：《数学的实践与认识》2011年第8期73-77,共5页Mathematics in Practice and Theory

基　　金：全国统计科研项目(2009LZ009)

摘　　要：采用Bayes分析方法,对完全随机缺失下配对试验设计数据进行统计分析,给出了参数的Bayes后验概率密度函数及参数的置信区间计算表达式,并对基于Bayes分析的假设检验功效进行了模拟.从模拟结果看,Bayes分析检验功效一致性地高于缺失数据配对删除的t检验方法,特别,在缺失数据相对较多或配对数据相关系数较低时,其表现出更大优势.Using Bayes analysis method, paired design with random missing data has been Analyzed. The parameters of Bayes posterior probability density function and parameters of the confidence interval calculation expression have also been deduced in this paper. At the same time, the power of test between Bayesian hypothesis test and the missing data paired deleting t-test has been compared by computer simulation, from the simulation results, Bayesian hypothesis test power consistently higher than the missing data paired deleting t- test, Especially, when the correlation coefficient is close to zero or the missing part hold a relatively large proportion, Bayes analysis method show a greater advantage.

关 键 词：配对试验 缺失数据 假设检验 BAYES统计 

分 类 号：O212.8[理学—概率论与数理统计]