Bayesian Structural Equation Modeling

Dr. Sam Manski, Dr. Wenjuan Ma, Dr. Frank Lawrence, CSTAT, Michigan State University

Bayesian statistical methods are becoming increasingly common due to their numerous advantages such as suitability for small sample sizes and ability to leverage past research by using informative prior distributions. These advantages extend to Structural Equation Modeling (SEM). In this presentation, we motivate the use of Bayesian methods in SEM and discuss Bayesian theory, the use of prior distributions, and assessing Bayesian performance. We then demonstrate these ideas and compare frequentist and Bayesian SEMs through a practical example in statistical software.