The Multi-Stage Gibbs Sampler

摘要: After two chapters of preparation on the slice and two-stage Gibbs samplers, respectively, we are now ready to envision entire picture for sampler. We describe general method in Section 10.1, whose theoretical properties less complete than special case (see 10.2): The defining difference between that sampler multi-stage version considered here is interleaving structure chain does not carry over. Some consequences fact individual subchains also Markov chains, Duality Principle Rao-Blackwellization hold some generality. None true here, case. Nevertheless, enjoys many optimality properties, still might be workhorse MCMC world. remainder this chapter deals with implementation considerations, connection important role Bayesian Statistics.