多变异分析的方差分量置信区间  

Confidence Intervals for Variance Components Based on Multi-Vari Analysis

作　　者：何桢[1] 刘晓亮[1] 

机构地区：[1]天津大学管理学院,天津300072

出　　处：《天津大学学报（社会科学版）》2006年第6期401-404,共4页Journal of Tianjin University：Social Sciences

基　　金：国家自然科学基金资助项目(70572044;70372062);教育部新世纪优秀人才资助计划(NCET-04-0240);天津市科技攻关资助项目(04310881R)

摘　　要：多变异分析将过程质量特征值的变异分为时间变异、产品间变异和产品内变异。产品间变异和时间变异的概率分布均为卡方随机变量的线性组合,而这些线性组合的分布表达式尚未求出,所以这两种变异的方差分量不能给出确切的置信区间。在介绍了求方差分量近似置信区间方法的基础上,将其与多变异分析的统计模型相结合,给出了产品间变异和时间变异分量的近似置信区间的求法。In multi-vari analysis （ MVA）, the variation sources of product/process quality characteristics are divided into time to time variation, piece to piece variation and within piece variation. But for the time to time and the piece to piece vafiance components, an expression for the distribution of their linear combination of chi-square random variables cannot be obtained. Thus, an exact confidence interval for them cannot be constructed. In this paper, approximate procedures are given. Based on the statistical model of MVA and the approximate procedures, the approximate confidence intervals of these two variance components are discussed.

关 键 词：多变异分析 方差分量 置信区间 修正的大样本方法 

分 类 号：F253.3[经济管理—国民经济]