文章基本信息

标题：Comment on Article by Jain and Neal
本地全文：下载
作者：David B. Dahl
期刊名称：Bayesian Analysis
印刷版ISSN：1931-6690
电子版ISSN：1936-0975
出版年度：2007
卷号：2
期号：3
页码：473--478
出版社：International Society for Bayesian Analysis
摘要：Sonia Jain and Radford Neal (JN) make a signicant contribution to the literature on Markov chain Monte Carlo (MCMC) sampling techniques for Dirichlet process mixture (DPM) models. The paper presents some very nice ideas and will be on my required reading list for students working with me. DPM models are widely used for Bayesian nonparametric analyses and ecient sampling techniques are essential for their routine application. Incremental samplers for nonconjugate DPM models, such as the Auxiliary Gibbs sampler in Neal (2000), are easily implemented and potentially very ecient. Unfortunately, these samplers can also have diculty mixing over the entire sample space and standard MCMC diagnostics may fail to indicate the problem. JN's paper represents a signicant advance by providing a non-incremental sampler for conditionally conjugate DPM models.