潜变量交互效应结构方程:分布分析方法  

Latent Interaction in Structural Equation Modeling:Distribution-analytic Approaches

作　　者：温忠麟[1] 吴艳[2] 侯杰泰[3] 

机构地区：[1]华南师范大学心理应用研究中心,广州510631 [2]广东外语外贸大学应用心理学系,广州510420 [3]香港中文大学教育心理系

出　　处：《心理学探新》2013年第5期409-414,共6页Psychological Exploration

基　　金：国家自然科学基金项目(31271116);教育部人文社会科学重点研究基地项目(11JJD190005);教育部人文社会科学研究青年基金项目(12YJC190031)

摘　　要：对于潜变量交互效应结构方程分析,目前应用较多的是乘积指标方法。分布分析方法国内还罕有应用,包括潜调节结构方程(LMS)方法和准极大似然(QML)方法。该研究以乘积指标方法的模型假设为参照,介绍了分布分析方法的模型假设。并简要叙述了LMS方法及其Mplus程序,QML方法及其QML程序。综合现有研究结果,总结出LMS和QML方法、无约束和约束方法的特点,从中可以看出各方法的优缺点,推荐了不同条件下合适的分析方法。Despite the many approaches to estimate interaction effects between two latent variables, historically product - indicator ap- proaches have been the most influential models. The present paper discusses less commonly used approaches in China, the distribution - analytic approaches, which are specialized alternatives for the estimation of non - linear structural equation models （SEM）. The distri- bution- analytic approaches include the Latent Moderated Structural Equations （LMS）approach and the Quasi -Maximum Likelihood （QML） approach. The assumptions of the distribution - analytic approaches are introduced and compared with those of the product - in- dicator approaches. The LMS approach and the related Mplus syntax, and the QML approach and QML syntax are briefed. The features of the LMS and QML approaches, and the unconstrained and constrained approaches are summarized in a table. The pros and cons of these approaches are compared and discussed with the appropriate analytical approach under different conditions being recommended.

关 键 词：潜变量 交互效应 潜调节结构方程 准极大似然 乘积指标 

分 类 号：B841.2[哲学宗教—基础心理学]