In this paper,we proposed a Bayesian analysis of finite mixtures in structural equation models(SEMs)using truncated Dirichlet process. The main idea of our modeling method is to reformulate the finite mixture SEMs as a single SEM,in which the truncated Dirichlet process priors are employed as the prior distributions of the involved unknown parameters.Given that,the Bayesian analysis can be implemented based on the Blocked Gibbs sampler.The proposed methodology has clear advantages over traditional mixture SEMs.First,it simultaneously determine the number of components in mixtures,estimates the unknown parameters in each component models,and provides the information of the Bayesian classification.Second,it can be easily implemented with the recently developed and freely available software WinBUGS,which leads practitioners avoid complicated derivation and non-trivial programming.A simulation study demonstrates that the performance of the proposed method is satisfactory.A study on a cocaine use data is provided to illustrate the methodology.……
