BIC extensions for order-constrained model selection

Joris Mulder*, A. E. Raftery

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

1 Downloads (Pure)

Abstract

The Schwarz or Bayesian information criterion (BIC) is one of the most widely used tools for model comparison in social science research. The BIC, however, is not suitable for evaluating models with order constraints on the parameters of interest. This article explores two extensions of the BIC for evaluating order-constrained models, one where a truncated unit information prior is used under the order-constrained model and the other where a truncated local unit information prior is used. The first prior is centered on the maximum likelihood estimate, and the latter prior is centered on a null value. Several analyses show that the order-constrained BIC based on the local unit information prior works better as an Occam’s razor for evaluating order-constrained models and results in lower error probabilities. The methodology based on the local unit information prior is implemented in the R package “BICpack” which allows researchers to easily apply the method for order-constrained model selection. The usefulness of the methodology is illustrated using data from the European Values Study.
Original languageEnglish
Number of pages28
JournalSociological Methods & Research
DOIs
Publication statusE-pub ahead of print - 2020

Fingerprint

EVS
model comparison
methodology
social science
Values

Keywords

  • BAYES FACTORS
  • Bayesian information criterion
  • European Values Study
  • HYPOTHESES
  • model selection
  • order constraints
  • truncated priors

Cite this

@article{8d1e8a57fc614aae8c35b2794b99909b,
title = "BIC extensions for order-constrained model selection",
abstract = "The Schwarz or Bayesian information criterion (BIC) is one of the most widely used tools for model comparison in social science research. The BIC, however, is not suitable for evaluating models with order constraints on the parameters of interest. This article explores two extensions of the BIC for evaluating order-constrained models, one where a truncated unit information prior is used under the order-constrained model and the other where a truncated local unit information prior is used. The first prior is centered on the maximum likelihood estimate, and the latter prior is centered on a null value. Several analyses show that the order-constrained BIC based on the local unit information prior works better as an Occam’s razor for evaluating order-constrained models and results in lower error probabilities. The methodology based on the local unit information prior is implemented in the R package “BICpack” which allows researchers to easily apply the method for order-constrained model selection. The usefulness of the methodology is illustrated using data from the European Values Study.",
keywords = "BAYES FACTORS, Bayesian information criterion, European Values Study, HYPOTHESES, model selection, order constraints, truncated priors",
author = "Joris Mulder and Raftery, {A. E.}",
year = "2020",
doi = "10.1177/0049124119882459",
language = "English",
journal = "Sociological Methods and Research",
issn = "0049-1241",
publisher = "SAGE Publications Inc.",

}

BIC extensions for order-constrained model selection. / Mulder, Joris; Raftery, A. E. .

In: Sociological Methods & Research, 2020.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - BIC extensions for order-constrained model selection

AU - Mulder, Joris

AU - Raftery, A. E.

PY - 2020

Y1 - 2020

N2 - The Schwarz or Bayesian information criterion (BIC) is one of the most widely used tools for model comparison in social science research. The BIC, however, is not suitable for evaluating models with order constraints on the parameters of interest. This article explores two extensions of the BIC for evaluating order-constrained models, one where a truncated unit information prior is used under the order-constrained model and the other where a truncated local unit information prior is used. The first prior is centered on the maximum likelihood estimate, and the latter prior is centered on a null value. Several analyses show that the order-constrained BIC based on the local unit information prior works better as an Occam’s razor for evaluating order-constrained models and results in lower error probabilities. The methodology based on the local unit information prior is implemented in the R package “BICpack” which allows researchers to easily apply the method for order-constrained model selection. The usefulness of the methodology is illustrated using data from the European Values Study.

AB - The Schwarz or Bayesian information criterion (BIC) is one of the most widely used tools for model comparison in social science research. The BIC, however, is not suitable for evaluating models with order constraints on the parameters of interest. This article explores two extensions of the BIC for evaluating order-constrained models, one where a truncated unit information prior is used under the order-constrained model and the other where a truncated local unit information prior is used. The first prior is centered on the maximum likelihood estimate, and the latter prior is centered on a null value. Several analyses show that the order-constrained BIC based on the local unit information prior works better as an Occam’s razor for evaluating order-constrained models and results in lower error probabilities. The methodology based on the local unit information prior is implemented in the R package “BICpack” which allows researchers to easily apply the method for order-constrained model selection. The usefulness of the methodology is illustrated using data from the European Values Study.

KW - BAYES FACTORS

KW - Bayesian information criterion

KW - European Values Study

KW - HYPOTHESES

KW - model selection

KW - order constraints

KW - truncated priors

U2 - 10.1177/0049124119882459

DO - 10.1177/0049124119882459

M3 - Article

JO - Sociological Methods and Research

JF - Sociological Methods and Research

SN - 0049-1241

ER -