2015/03/18 by Daniel Turek, Turek, Daniel, Perry de Valpine +5 · 1 citation
Mathematics · #Computation (stat.CO) #FOS: Computer and information sciences #Markov Chains and Monte Carlo Methods #Statistical Methods and Bayesian Inference #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.1503.05621
openalex publication_date 2015/03/18 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Markov chain Monte Carlo (MCMC) sampling is an important and commonly used\ntool for the analysis of hierarchical models. Nevertheless, practitioners\ngenerally have two options for MCMC: utilize existing software that generates a\nblack-box "one size fits all" algorithm, or the challenging (and time\nconsuming) task of implementing a problem-specific MCMC algorithm. Either\nchoice may result in inefficient sampling, and hence researchers have become\naccustomed to MCMC runtimes on the order of days (or longer) for large models.\nWe propose an automated procedure to determine an efficient MCMC algorithm for\na given model and computing platform. Our procedure dynamically determines\nblocks of parameters for joint sampling that result in efficient sampling of\nthe entire model. We test this procedure using a diverse suite of example\nmodels, and observe non-trivial improvements in MCMC efficiency for many\nmodels. Our procedure is the first attempt at such, and may be generalized to a\nbroader space of MCMC algorithms. Our results suggest that substantive\nimprovements in MCMC efficiency may be practically realized using our automated\nblocking procedure, or variants thereof, which warrants additional study and\napplication.\n