摘要
在传统元分析中,每个研究只能提取一个效果量,以满足各效果量相互独立的条件。若出现单个研究多效果量的情况,就需要采用多元元分析技术予以处理。多元元分析主要包括多元线性模型法和多元整合法,其中,多元整合法是应用最为广泛的一种方法,可较为有效地解决多效果量非独立性问题,还可借助传统元分析的固定效应模型和混合效应模型进行统计分析。但多元整合法在应用中还存在着统计软件的开发问题、同质性Q检验显著性的稳定问题、不同质混合效果量的同质性检验的可行性问题。
In traditional meta-analysis, only a single effect size estimate can be produced in each study to make each effect size estimate be independent from each other. If there are multiple effect size estimates in a single study, multivariate meta-analysis technique is needed to deal with the data. Multivariate meta-analysis includes multivariate linear modeling and multivariate integrative method, in which multivariate integrative method is one of the widest used approaches. Multivariate integrative method can solve the problem of multiple dependent effect sizes effectively and can also do the statistical analysis by fixed effects models and mixed effects models. However, some problems still exist in the appliance of multivariate integrative method, such as the development of the Statistical software, the stability of the significance of homogeneity Q test and the feasibility of homogeneity test of heterogeneous synthetic effect sizes.
出处
《心理科学进展》
CSSCI
CSCD
北大核心
2010年第3期505-510,共6页
Advances in Psychological Science
关键词
多元整合法
多元元分析
效果量
非独立性
混合效果量
multivariate integrative method
multivariate meta-analysis
effect size
dependence
synthetic effect size
