Investigation of the widely applicable Bayesian information criterion

DOI10.1007/S11222-016-9657-YMaRDI QIDQ2361460zbMATH OpenOpenAlexFDO

Authors Nial Friel, James P. McKeone, Chris J. Oates, A. N. Pettitt

Publication date 30 June 2017

Published in Statistics and Computing (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1501.05447

zbMATH Keywords

marginal likelihoodevidencepower posteriorswidely applicable Bayesian information criterion

Mathematics Subject Classification ID

Computational methods for problems pertaining to statistics (62-08) Bayesian inference (62F15)

Abstract: The widely applicable Bayesian information criterion (WBIC) is a simple and fast approximation to the model evidence that has received little practical consideration. WBIC uses the fact that the log evidence can be written as an expectation, with respect to a powered posterior proportional to the likelihood raised to a power $t^{\ast }in(0,1)$, of the log deviance. Finding this temperature value $t^{\ast }$ is generally an intractable problem. We find that for a particular tractable statistical model that the mean squared error of an optimally-tuned version of WBIC with correct temperature $t^{\ast }$ is lower than an optimally-tuned version of thermodynamic integration (power posteriors). However in practice WBIC uses the a canonical choice of $t=1/log(n)$. Here we investigate the performance of WBIC in practice, for a range of statistical models, both regular models and singular models such as latent variable models or those with a hierarchical structure for which BIC cannot provide an adequate solution. Our findings are that, generally WBIC performs adequately when one uses informative priors, but it can systematically overestimate the evidence, particularly for small sample sizes.

Recommendations

• A case study of the widely applicable Bayesian information criterion and its optimality
• A widely applicable Bayesian information criterion
• Understanding predictive information criteria for Bayesian models
• A Bayesian information criterion for singular models. With discussion and authors' reply
• Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC

Cites work

• scientific article; zbMATH DE number 3144053 (Why is no real title available?)
• scientific article; zbMATH DE number 597911 (Why is no real title available?)
• A Bayesian information criterion for singular models. With discussion and authors' reply
• A New Approach to Numerical Integration
• A Reference Bayesian Test for Nested Hypotheses and its Relationship to the Schwarz Criterion
• A case study of the widely applicable Bayesian information criterion and its optimality
• A widely applicable Bayesian information criterion
• An adaptive scheduling scheme for calculating Bayes factors with thermodynamic integration using Simpson's rule
• Bayesian Information Criterion for Censored Survival Models
• Efficient approximations for the marginal likelihood of Bayesian networks with hidden variables
• Estimating Bayes factors via thermodynamic integration and population MCMC
• Estimating the dimension of a model
• Improving power posterior estimation of statistical evidence
• Marginal Likelihood Estimation via Power Posteriors
• Nested sampling for general Bayesian computation
• Reversible jump Markov chain Monte Carlo computation and Bayesian model determination
• Simulating normalizing constants: From importance sampling to bridge sampling to path sampling
• Statistical Mechanics of Fluid Mixtures
• Thermodynamic Bayesian model comparison
• Understanding predictive information criteria for Bayesian models

Cited in

(8)

• Generalized information criteria for Bayes decisions
• A case study of the widely applicable Bayesian information criterion and its optimality
• Bayesian clustering of skewed and multimodal data using geometric skewed normal distributions
• Clustering ultrasonic waves propagation time: a hierarchical polynomial semiparametric approach
• Higher order equivalence of Bayes cross validation and WAIC
• A widely applicable Bayesian information criterion
• Bayesian model-based clustering for longitudinal ordinal data
• Bayesian predictive information criterion for the evaluation of hierarchical Bayesian and empirical Bayes models

This page was built for publication: Investigation of the widely applicable Bayesian information criterion

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2361460)