Bayesian model selection for multilevel mediation models

Mediation analysis is often used to explore the complex relationship between two variables through a third mediating variable. This paper aims illustrate performance of deviance information criterion, pseudo‐Bayes factor, and Watanabe–Akaike criterion in selecting appropriate multilevel mediation model. Our focus will be on comparing conditional criteria (given random effects) versus marginal (averaged over this respect. Most previous work models fails report poor behavior criteria. We demonstrate here superiority version selection their counterpart mediated longitudinal settings simulation studies via an application data from Longitudinal Aging Study Amsterdam study. In addition, we usefulness our self‐written R function for models.