结构方程模型下多因子非同质测量合成分数的信度估计  

Using the Structural Equation Modeling Approach to Estimate Multi-factor Non-congeneric Composite Reliability

作　　者：屠金路[1] 王庭照[1,2] 金瑜[3] 

机构地区：[1]陕西师范大学心理学系,西安710062 [2]南京航空航天大学管理科学与工程系,南京210016 [3]华东师范大学特殊教育学系,上海200062

出　　处：《心理科学》2010年第3期666-669,共4页Journal of Psychological Science

基　　金：陕西师范大学校级重点研究支持;陕西省2009年基础教育科研重大招标课题(zdk10910-1)支持

摘　　要：在Dillon提出的单因子同质测量合成分数信度估计公式的基础上,本文尝试推导出了一个结构方程模型下多因子非同质测量合成分数的信度估计公式,并进一步对该公式在权重不等、误差相关等条件下的扩展进行了推导和讨论。同时,本文采用一组多元正态的模拟数据,对推导出的信度估计公式进行了检验。结果发现,该公式相对Raykov提出的两因子非同质测量合成分数信度估计方法和α系数法具有较大的优越性。Based on the congeneric scale reliability formula advanced by Dillon, a multi-factor non-congeneric composite reliability formula was introduced according to the structural equation modeling, and could be extended to estimate composite reliability when the error was corrected and the items were with unweight. With a set of simulated data, the composite reliability was computed. The new formula was found to be consisent with Raykov＇ s. The formula more accurately estimated composite reliability than coefficient alpha

关 键 词：结构方程模型 多因子非同质测量 a系数 合成分数信度 

分 类 号：B841.7[哲学宗教—基础心理学]