典型相关分析与结构方程模型方法的比较研究  

A Comparative Study of Canonical Correlation Analysis and Structural Equation Modeling

作　　者：刘怡然 安奉钧[2] Liu Yiran;An Fengjun(School of Public Security Management,People’s Public Security University of China,Beijing 100038,China;School of Public Management,Hebei University of Economics and Business,Shijiazhuang 050061,China)

机构地区：[1]中国人民公安大学公安管理学院,北京100038 [2]河北经贸大学公共管理学院,石家庄050061

出　　处：《统计与决策》2024年第10期40-45,共6页Statistics & Decision

基　　金：四川省智慧警务与国家安全风险治理重点实验室重点项目(ZHZZZD2302)。

摘　　要：典型相关分析(CCA)和结构方程模型(SEM)的应用日益广泛,但由于二者存在较大的相似性而常使研究者在使用过程中面临选择困难等问题,因而准确理解和区分二者之间的差异至关重要。文章基于SPSS和AMOS操作环境对典型相关分析和结构方程模型在函数式、基本图形、二阶因素、中介效应及使用条件等方面进行系统性比较,研究结果表明:(1)CCA可以进行线性组合计算,直接计算一个潜变量与另一组显变量的关系,能有效处理二阶因素计算问题。(2)SEM可以同时考虑多个潜变量之间的关系、计算和呈现误差方差和残差、准确地计算和展示中介效应、运用辅助性标准来判断模型适配度并通过调整变量之间的逻辑联系来修正模型,其输出结果全面而精确。(3)CCA适用于含有两个潜变量的简单模型,而SEM适用于含有多个潜变量的复杂模型。在特定条件下,可以结合使用CCA和SEM,或用CCA代替SEM。Canonical Correlation Analysis(CCA)and Structural Equation Modeling(SEM)are more and more widely used,but because of their similarities,researchers often face difficulties in choosing between them.Thus,it is of great significance to distinguish the relationship between the two.Based on the operating environment of SPSS and AMOS,this paper makes a system-atic comparison from the perspectives of functions,fundamental graphics,second order factors,mediating effects and using condi-tions between CCA and SEM.The results go as below:(1)CCA can perform linear combination calculation,directly calculating the relationship between a latent variable and another group of manifest variables,and can effectively deal with the problem of sec-ond-order factor calculation.(2)SEM can show the relationships of multiple latent variables simultaneously,calculate and present error variances and residual,accurately calculate and show mediating effect,use auxiliary criteria to judge the fitness of the model,and modify the model by adjusting the logical relationship between the variables.Its output results are more comprehensive and accurate.(3)CCA is suitable for simple models with two latent variables,while SEM is suitable for complex models with multiple latent variables.Under certain conditions,CCA and SEM can be combined,or CCA can be used instead of SEM.

关 键 词：典型相关分析 结构方程模型 潜变量 负荷量 适配度 

分 类 号：O29[理学—应用数学]