Automated Parameter Blocking for Efficient Markov Chain Monte Carlo Sampling

• Activity Number: 133
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Daniel Turek*
• Companies: UC Berkeley
• Keywords: MCMC ; Metropolis-Hastings ; sampling efficiency ; block sampling

Abstract

Practitioners are generally restricted to two options for carrying out MCMC analyses: utilize existing software that generates a black-box "one size fits all" MCMC algorithm, which generally is not optimized to the problem at hand, or the challenging (and time consuming) task of implementing a problem-specific MCMC algorithm. Either choice may result in inefficient MCMC sampling, and hence researchers have become accustomed to MCMC runtimes on the order of days for large hierarchical models. We propose an automated procedure to determine an efficient MCMC algorithm for a given model and data set. The procedure dynamically determines blocks of model parameters for joint sampling that result in efficient sampling of the entire model. Our procedure is tested using a suite of example models and compared against statically-specified MCMC algorithms. Automated blocking produces non-trivial improvements in MCMC efficiency for many models. Our results suggest that substantive improvements in MCMC efficiency may be practically realized using our automated blocking procedure, or variants thereof, which warrants additional study and application.