结构方程模型中拟合指数的运用与比较  

THE COMPARISION OF GOODNESS INDEX OF STRUCTURAL EQUATION MODEL

作　　者：王长义[1] 王大鹏[1] 赵晓雯[1] 方庆伟[1] 刘艳[1] 

机构地区：[1]哈尔滨医科大学公共卫生学院卫生统计学教研室,黑龙江100081

出　　处：《现代预防医学》2010年第1期7-9,共3页Modern Preventive Medicine

基　　金：黑龙江省自然科学基金(D2005-55);黑龙江省教育厅项目(11521142);哈尔滨市科技攻关项目(2008RFQXS086)

摘　　要：结构方程模型是一种建立、估计和检验因果关系模型的方法,近二十年来在心理学、社会学、行为科学以及医学等领域得到越来越广泛的应用,通常利用拟合指数对模型的拟合效果进行评价。然而,结构方程模型的拟合指数有40多种,不同的拟合指数反映出的拟合效果并非完全一致,因此拟合指数性能的可靠性成为应用者比较关心的话题。本文通过文献查阅,较为系统地总结了结构方程模型常用拟合指数的分类及其特征,通过对比分析发现指数NNFI和RMSEA相对值得信赖,CFI、RNI和Mc具有一定的参考价值,而在用ML或GLS法进行参数估计时SRMR则表现出了较好的效果。Structural equation model （SEM）,which is a method of building,estamating and testing causality modeling,has been being used widely in the fields of psychology,sociology,behavior science and medicine. Goodness index is usually used for evaluating the goodness effect of models,but there are more than 40 indexs for SEM,and the results of different ones are not same. So which indexes are credible becomes a concerned topic for users. On the base of collecting many literatures,we summarized systematically the category and characters of indexes of SEM. By the method of comparing analysis,we find NNFI and RMSEA are credible,the second is CFI,RNI and Mc. When estamating parameters with ML or GLS,we find SRMR has better effect.

关 键 词：结构方程模型 拟合指数 拟合效果 

分 类 号：R195.1[医药卫生—卫生统计学]