Harris recurrence of Metropolis-within-Gibbs and trans-dimensional Markov chains

Abstract

A $\varphi$-irreducible and aperiodic Markov chain with stationary probability distribution will converge to its stationary distribution from almost all starting points. The property of Harris recurrence allows us to replace ``almost all'' by ``all,'' which is potentially important when running Markov chain Monte Carlo algorithms. Full-dimensional Metropolis--Hastings algorithms are known to be Harris recurrent. In this paper, we consider conditions under which Metropolis-within-Gibbs and trans-dimensional Markov chains are or are not Harris recurrent. We present a simple but natural two-dimensional counter-example showing how Harris recurrence can fail, and also a variety of positive results which guarantee Harris recurrence. We also present some open problems. We close with a discussion of the practical implications for MCMC algorithms.

Published at http://dx.doi.org/10.1214/105051606000000510 in the Annals of Applied Probability (http://www.imstat.org/aap/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (1)

• General state space Markov chains and MCMC algorithms

Cited by in corpus (14)

• Importance sampling type estimators based on approximate marginal MCMC
• Bayesian Variable Selection for Probit Mixed Models Applied to Gene Selection
• Gradient Scan Gibbs Sampler: an efficient algorithm for high-dimensional Gaussian distributions
• Importance sampling correction versus standard averages of reversible MCMCs in terms of the asymptotic variance
• Network Inference Using Steady State Data and Goldbeter-Koshland Kinetics
• Covid19 Reproduction Number: Credibility Intervals by Blockwise Proximal Monte Carlo Samplers
• Kalman filtering and smoothing for linear wave equations with model error
• A Quasi-Bayesian Perspective to Online Clustering
• Slice Sampling for General Completely Random Measures
• Correlated pseudo-marginal Metropolis-Hastings using quasi-Newton proposals
• Logit unfolding choice models for binary data
• Temporal Topic Analysis with Endogenous and Exogenous Processes
• Monte Carlo methods for light propagation in biological tissues
• Bayesian spline-based hidden Markov models with applications to actimetry data and sleep analysis