Bayesian Analysis of Mixture Normal Model via Equi-Energy Sampler and Improved Metropolis-Hastings Algorithm
Wei Shao, Guoqing Zhao, Feifei Shao · 2015
Mixture normal model provides a convenient and flexible probabilistic representation of heterogeneous data, and the estimation of parameters received considerable attention in recent years. In this paper, we propose a Bayesian analysis of mixture normal model. Because the the posterior probability density function is too complicated to be used to draw samples directly using standard Markov Chain Monte Carlo method, we use two method, the Improved Metropolis-Hastings algorithm and Equi-energy sampler, to conquer the drawback. We show by numerical simulations that both Equi-energy sampler and Improved Metropolis-Hastings algorithm outperform the standard Metropolis-Hastings algorithm.