Bayesian Clustering for Ordinal Data Based on Finite Mixture Models of Latent Variables

Based on finite mixture models of latent variables, we propose a Bayesian clustering method for ordinal data. EM algorithm is employed to compute the estimates of model parameters, BIC criterion is adopted to determine the number of clusters, and an analogue procedure of Bayesian discrimination is used to classify each observation. The results of a simulation study show that the method works well, and the computing efficiency is acceptable for a dataset of moderate size.