贝叶斯结构方程模型的原理及应用

The Principle of Bayesian Structural Equation Model and Its Application in Psychology

贝叶斯结构方程模型同时兼容了结构方程模型考虑测量误差以及贝叶斯估计方法可以整合先验信息的优点,在处理传统的结构方程模型针对交叉载荷参数、残差协方差元素严格限制为0等假设时具有一定的优势,且更易处理小样本和复杂模型的情况。但该方法在国内心理学领域的应用仍相对较少。本文主要介绍贝叶斯结构方程模型的原理、先验设置、模型收敛判断和模型评估等问题,并以具体示例展示如何在Mplus软件中使用该模型处理交叉载荷的问题,希望可以帮助更多研究者认识和了解贝叶斯结构方程模型,并使用该模型解决一些实际的研究问题。

Bayesian structural equation model offers a methodology to incorporate researcher’s theories and prior beliefs by utilizing informative priors for parameters with strict restrictions,such as assuming no crossloadings.This model combines the benefits of structural equation model,which accounts for measurement error,with the advantages of Bayesian estimation method,which integrates prior information.It is easier to deal with small samples and complex models.Furthermore,it provides advantages by relaxing strict assumptions found in traditional structural equation models,such as those regarding cross-loading parameters and the diagonal residual covariance matrix.Nevertheless,the utilization of the Bayesian structural equation model in the field of psychology in China remains relatively limited.The primary objective of this study is to introduce the principles,utilization of priors,assessment of model convergence,model evaluation,and other pertinent aspects of the Bayesian structural equation model.Additionally,it illustrates the application of the Bayesian structural equation model in addressing potential crossloading parameters using the Mplus software,accompanied by an example.The ultimate aim is to enhance researchers’understanding of the Bayesian structural equation model and its potential for resolving practical research problems.