Title: Semiparametric Bayesian analysis for longitudinal mixed effects models with non-normal AR(1) errors
Authors: Junshan Shen - Capital Univeristy of Economics and Business (China)
Jin Yang - The Hong Kong polytechnic University (Hong Kong)
Catherine Liu - The Hong Kong Polytechnic University (Hong Kong) [presenting]
Abstract: The focus is on Bayesian inference on the longitudinal mixed effects model with non-normal AR(1) errors. We model the nonparametric zero-mean noise in the autoregression residual by the Dirichlet process (DP) mixture model. Applying the empirical likelihood tool, an adjusted sampler based on the Polya urn representation of DP is proposed to incorporate information of the moment constraints of the mixing distribution. Gibbs sampling algorithm based on the adjusted sampler is proposed to approximate the posterior distributions under DP priors. The proposed method can be easily extended to deal with other moment constraints owing to the wide application background of empirical likelihood. Simulation studies evaluate the performance of the proposed method. Our method is illustrated in analysis of a longitudinal data set from a psychiatric study.