A Comparison of Bayes Factor Approximation Methods Including Two New Methods

Kenneth A. Bollen, Surajit Ray, Jane R Zavisca, Jeffrey J. Harden

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

Bayes factors (BFs) play an important role in comparing the fit of statistical models. However, computational limitations or lack of an appropriate prior sometimes prevent researchers from using exact BFs. Instead, it is approximated, often using the Bayesian Information Criterion (BIC) or a variant of BIC. The authors provide a comparison of several BF approximations, including two new approximations, the Scaled Unit Information Prior Bayesian Information Criterion (SPBIC) and Information matrix-based Bayesian Information Criterion (IBIC). The SPBIC uses a scaled unit information prior that is more general than the BIC's unit information prior, and the IBIC utilizes more terms of approximation than the BIC. Through simulation, the authors show that several measures perform well in large samples, that performance declines in smaller samples, and that SPBIC and IBIC provide improvement to existing measures under some conditions, including small sample sizes. The authors illustrate the use of the fit measures with the crime data of Ehrlich and then conclude with recommendations for researchers.

Original languageEnglish (US)
Pages (from-to)294-324
Number of pages31
JournalSociological Methods and Research
Volume41
Issue number2
DOIs
StatePublished - 2012

Fingerprint

offense
simulation
lack
performance

Keywords

  • Bayes factor
  • empirical Bayes
  • information criterion
  • Laplace approximation
  • model selection
  • scaled unit information prior
  • variable selection

ASJC Scopus subject areas

  • Social Sciences (miscellaneous)
  • Sociology and Political Science

Cite this

A Comparison of Bayes Factor Approximation Methods Including Two New Methods. / Bollen, Kenneth A.; Ray, Surajit; Zavisca, Jane R; Harden, Jeffrey J.

In: Sociological Methods and Research, Vol. 41, No. 2, 2012, p. 294-324.

Research output: Contribution to journalArticle

Bollen, Kenneth A. ; Ray, Surajit ; Zavisca, Jane R ; Harden, Jeffrey J. / A Comparison of Bayes Factor Approximation Methods Including Two New Methods. In: Sociological Methods and Research. 2012 ; Vol. 41, No. 2. pp. 294-324.
@article{49768b89768845a1b408b8bacbc093aa,
title = "A Comparison of Bayes Factor Approximation Methods Including Two New Methods",
abstract = "Bayes factors (BFs) play an important role in comparing the fit of statistical models. However, computational limitations or lack of an appropriate prior sometimes prevent researchers from using exact BFs. Instead, it is approximated, often using the Bayesian Information Criterion (BIC) or a variant of BIC. The authors provide a comparison of several BF approximations, including two new approximations, the Scaled Unit Information Prior Bayesian Information Criterion (SPBIC) and Information matrix-based Bayesian Information Criterion (IBIC). The SPBIC uses a scaled unit information prior that is more general than the BIC's unit information prior, and the IBIC utilizes more terms of approximation than the BIC. Through simulation, the authors show that several measures perform well in large samples, that performance declines in smaller samples, and that SPBIC and IBIC provide improvement to existing measures under some conditions, including small sample sizes. The authors illustrate the use of the fit measures with the crime data of Ehrlich and then conclude with recommendations for researchers.",
keywords = "Bayes factor, empirical Bayes, information criterion, Laplace approximation, model selection, scaled unit information prior, variable selection",
author = "Bollen, {Kenneth A.} and Surajit Ray and Zavisca, {Jane R} and Harden, {Jeffrey J.}",
year = "2012",
doi = "10.1177/0049124112452393",
language = "English (US)",
volume = "41",
pages = "294--324",
journal = "Sociological Methods and Research",
issn = "0049-1241",
publisher = "SAGE Publications Inc.",
number = "2",

}

TY - JOUR

T1 - A Comparison of Bayes Factor Approximation Methods Including Two New Methods

AU - Bollen, Kenneth A.

AU - Ray, Surajit

AU - Zavisca, Jane R

AU - Harden, Jeffrey J.

PY - 2012

Y1 - 2012

N2 - Bayes factors (BFs) play an important role in comparing the fit of statistical models. However, computational limitations or lack of an appropriate prior sometimes prevent researchers from using exact BFs. Instead, it is approximated, often using the Bayesian Information Criterion (BIC) or a variant of BIC. The authors provide a comparison of several BF approximations, including two new approximations, the Scaled Unit Information Prior Bayesian Information Criterion (SPBIC) and Information matrix-based Bayesian Information Criterion (IBIC). The SPBIC uses a scaled unit information prior that is more general than the BIC's unit information prior, and the IBIC utilizes more terms of approximation than the BIC. Through simulation, the authors show that several measures perform well in large samples, that performance declines in smaller samples, and that SPBIC and IBIC provide improvement to existing measures under some conditions, including small sample sizes. The authors illustrate the use of the fit measures with the crime data of Ehrlich and then conclude with recommendations for researchers.

AB - Bayes factors (BFs) play an important role in comparing the fit of statistical models. However, computational limitations or lack of an appropriate prior sometimes prevent researchers from using exact BFs. Instead, it is approximated, often using the Bayesian Information Criterion (BIC) or a variant of BIC. The authors provide a comparison of several BF approximations, including two new approximations, the Scaled Unit Information Prior Bayesian Information Criterion (SPBIC) and Information matrix-based Bayesian Information Criterion (IBIC). The SPBIC uses a scaled unit information prior that is more general than the BIC's unit information prior, and the IBIC utilizes more terms of approximation than the BIC. Through simulation, the authors show that several measures perform well in large samples, that performance declines in smaller samples, and that SPBIC and IBIC provide improvement to existing measures under some conditions, including small sample sizes. The authors illustrate the use of the fit measures with the crime data of Ehrlich and then conclude with recommendations for researchers.

KW - Bayes factor

KW - empirical Bayes

KW - information criterion

KW - Laplace approximation

KW - model selection

KW - scaled unit information prior

KW - variable selection

UR - http://www.scopus.com/inward/record.url?scp=84864958818&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84864958818&partnerID=8YFLogxK

U2 - 10.1177/0049124112452393

DO - 10.1177/0049124112452393

M3 - Article

AN - SCOPUS:84864958818

VL - 41

SP - 294

EP - 324

JO - Sociological Methods and Research

JF - Sociological Methods and Research

SN - 0049-1241

IS - 2

ER -