Predictive bayesian model selection

Tomohiro Ando

Research output: Contribution to journalArticle

40 Citations (Scopus)

Abstract

We investigate the problem of evaluating the goodness of the predictive distributions of Bayesian models. Recently, deviance information criteria (DIC) has been extensively employed in various study areas to evaluate the Bayesian models, thanks to its simplicity of calculation from the posterior simulation outputs. Unfortunately, it is known that DIC often selects overfitted models. In this paper, we develop a new criterion which can be calculated easily from posterior outputs under the model misspecification situation. The proposed criterion is developed as an estimator of the posterior mean of the expected likelihood and is robust to improper priors. Monte Carlo simulations are conducted to investigate the properties of the proposed criteria.

Original languageEnglish
Pages (from-to)13-38
Number of pages26
JournalAmerican Journal of Mathematical and Management Sciences
Volume31
Issue number1-2
Publication statusPublished - 2011

Fingerprint

Bayesian Model Selection
Predictive Model
Deviance Information Criterion
Bayesian Model
Improper Prior
Posterior Mean
Model Misspecification
Predictive Distribution
Output
Simplicity
Likelihood
Monte Carlo Simulation
Estimator
Evaluate
Model selection
Bayesian model
Simulation
Deviance
Information criterion
Model

Keywords

  • Effective number of parameters
  • Empirical bayes markov chain monte carlo
  • Hierarchical bayesian model
  • Model misspecification

ASJC Scopus subject areas

  • Business, Management and Accounting(all)
  • Applied Mathematics

Cite this

Predictive bayesian model selection. / Ando, Tomohiro.

In: American Journal of Mathematical and Management Sciences, Vol. 31, No. 1-2, 2011, p. 13-38.

Research output: Contribution to journalArticle

Ando, Tomohiro. / Predictive bayesian model selection. In: American Journal of Mathematical and Management Sciences. 2011 ; Vol. 31, No. 1-2. pp. 13-38.
@article{176ce4742d7e4e82b397bc3d91debe6a,
title = "Predictive bayesian model selection",
abstract = "We investigate the problem of evaluating the goodness of the predictive distributions of Bayesian models. Recently, deviance information criteria (DIC) has been extensively employed in various study areas to evaluate the Bayesian models, thanks to its simplicity of calculation from the posterior simulation outputs. Unfortunately, it is known that DIC often selects overfitted models. In this paper, we develop a new criterion which can be calculated easily from posterior outputs under the model misspecification situation. The proposed criterion is developed as an estimator of the posterior mean of the expected likelihood and is robust to improper priors. Monte Carlo simulations are conducted to investigate the properties of the proposed criteria.",
keywords = "Effective number of parameters, Empirical bayes markov chain monte carlo, Hierarchical bayesian model, Model misspecification",
author = "Tomohiro Ando",
year = "2011",
language = "English",
volume = "31",
pages = "13--38",
journal = "American Journal of Mathematical and Management Sciences",
issn = "0196-6324",
publisher = "American Sciences Press Inc.",
number = "1-2",

}

TY - JOUR

T1 - Predictive bayesian model selection

AU - Ando, Tomohiro

PY - 2011

Y1 - 2011

N2 - We investigate the problem of evaluating the goodness of the predictive distributions of Bayesian models. Recently, deviance information criteria (DIC) has been extensively employed in various study areas to evaluate the Bayesian models, thanks to its simplicity of calculation from the posterior simulation outputs. Unfortunately, it is known that DIC often selects overfitted models. In this paper, we develop a new criterion which can be calculated easily from posterior outputs under the model misspecification situation. The proposed criterion is developed as an estimator of the posterior mean of the expected likelihood and is robust to improper priors. Monte Carlo simulations are conducted to investigate the properties of the proposed criteria.

AB - We investigate the problem of evaluating the goodness of the predictive distributions of Bayesian models. Recently, deviance information criteria (DIC) has been extensively employed in various study areas to evaluate the Bayesian models, thanks to its simplicity of calculation from the posterior simulation outputs. Unfortunately, it is known that DIC often selects overfitted models. In this paper, we develop a new criterion which can be calculated easily from posterior outputs under the model misspecification situation. The proposed criterion is developed as an estimator of the posterior mean of the expected likelihood and is robust to improper priors. Monte Carlo simulations are conducted to investigate the properties of the proposed criteria.

KW - Effective number of parameters

KW - Empirical bayes markov chain monte carlo

KW - Hierarchical bayesian model

KW - Model misspecification

UR - http://www.scopus.com/inward/record.url?scp=84868296532&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84868296532&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84868296532

VL - 31

SP - 13

EP - 38

JO - American Journal of Mathematical and Management Sciences

JF - American Journal of Mathematical and Management Sciences

SN - 0196-6324

IS - 1-2

ER -