A Comparison of the Impact of Informative Priors for the Bayesian Discontinuous Piecewise Growth Mixture Model

• In Event: Bayesian Estimation of Structural Equation Models

Abstract

When using Bayesian analysis, researchers should carefully specify the prior for accurate estimation. This simulation study was designed to explore the accuracy of estimates by specifying the prior placed on latent classifications and growth parameters within the context of discontinuous piecewise growth mixture models. The simulation conditions are considered such as sample size, the mixture class proportions, the prior for growth parameters, and mixture class proportion. According to the results, even if the variance hyperparameter is set high, if the mean hyperparameter is close to the parameter value, it is found that estimates were estimated well. Thus, the mixture class proportion estimates recovered well when the informative prior distribution for the mixture class proposition was defined.

Authors

• Yeheun Baek, Chungnam National University

• Jungmin Lee, National Youth Policy Institute

• Hyewon Chung, Chungnam National University