摘要
"组内相关系数"正越来越多地被用于自然科学与社会科学诸领域,但国内外应用者对其定义与估计方法的理解尚有不足.其名称源于将"皮尔逊积矩相关"与对称表结合构成配对估计量的经典定义.而费希尔基于组间方差比重的新定义得益于哈里斯对配对估计量的简化.新定义在平衡数据下可由ANOVA法估计且与配对估计量渐近相等,故两种定义被统称为组内相关系数.在非平衡数据下有9个估计量可供选择,包括6个加权配对和3个方差成分类估计量.应用中需按观察变量是否符合正态分布假设等原则加以选择.本研究例解了方差成分类估计量的Stata命令.
Intra-class correlation coefficient （ICC） is increasingly being used in various fields of natural science and social science. Unfortunately, there＇s inadequate understanding of its definition and estimation methods in the appli- cation. It＇s very name origins from its classical definition, in which a pair-wise estimator （PE） was constructed through calculating Pearson Product-moment Correlation Coefficient in a symmetric table Benefited from a simplifica- tion of pair-wise estimator by Harris, Fisher proposed a new definition based on the percentage of between-group variance, which in balanced data can be estimated by ANOVA and is asymptomatieally equivalent to the PE. Thus, both of these two definitions were referred to as ICC. For unbalanced data, nine ICC estimators are available, inclu- ding six weighted-pairwise （WP） estimators and three variance-components （VC） estimators. In the application, we should choose the one best fit using several criteria, such as the normality of variables, etc. The calculations of VC estimators with a couple of Stata commands are illustrated.
出处
《浙江大学学报（理学版）》
CAS
CSCD
2013年第5期509-515,共7页
Journal of Zhejiang University（Science Edition）
基金
国家自然科学基金青年项目(70903056)
关键词
组内相关系数
配对估计量
方差成分
非平衡数据
STATA
intraclass correlation coefficient
pair-wise estimator
variance components
unbalanced data
Stata
