Prior sensitivity analysis in default Bayesian structural equation modeling

S.J. van Erp, J. Mulder, Daniel L. Oberski

Research output: Contribution to journalArticleScientificpeer-review

74 Downloads (Pure)

Abstract

Bayesian structural equation modeling (BSEM) has recently gained popularity because it enables researchers to fit complex models and solve some of the issues often encountered in classical maximum likelihood estimation, such as nonconvergence and inadmissible solutions. An important component of any Bayesian analysis is the prior distribution of the unknown model parameters. Often, researchers rely on default priors, which are constructed in an automatic fashion without requiring substantive prior information. However, the prior can have a serious influence on the estimation of the model parameters, which affects the mean squared error, bias, coverage rates, and quantiles of the estimates. In this article, we investigate the performance of three different default priors: noninformative improper priors, vague proper priors, and empirical Bayes priors-with the latter being novel in the BSEM literature. Based on a simulation study, we find that these three default BSEM methods may perform very differently, especially with small samples. A careful prior sensitivity analysis is therefore needed when performing a default BSEM analysis. For this purpose, we provide a practical step-by-step guide for practitioners to conducting a prior sensitivity analysis in default BSEM. Our recommendations are illustrated using a well-known case study from the structural equation modeling literature, and all code for conducting the prior sensitivity analysis is available in the online supplemental materials.

Original languageEnglish
Pages (from-to)363-388
JournalPsychological Methods
Volume23
Issue number2
DOIs
Publication statusPublished - 2018

Keywords

  • Bayesian
  • CLASS SEPARATION
  • COMPONENTS
  • EMPIRICAL BAYES
  • FRAMEWORK
  • FREQUENTIST
  • INFERENCE
  • LINEAR MIXED MODELS
  • PARAMETERS
  • PRIOR DISTRIBUTIONS
  • SELECTION
  • default priors
  • sensitivity analysis
  • structural equation models

Cite this

@article{b536332bbca54e10876cdc6d901fe9bf,
title = "Prior sensitivity analysis in default Bayesian structural equation modeling",
abstract = "Bayesian structural equation modeling (BSEM) has recently gained popularity because it enables researchers to fit complex models and solve some of the issues often encountered in classical maximum likelihood estimation, such as nonconvergence and inadmissible solutions. An important component of any Bayesian analysis is the prior distribution of the unknown model parameters. Often, researchers rely on default priors, which are constructed in an automatic fashion without requiring substantive prior information. However, the prior can have a serious influence on the estimation of the model parameters, which affects the mean squared error, bias, coverage rates, and quantiles of the estimates. In this article, we investigate the performance of three different default priors: noninformative improper priors, vague proper priors, and empirical Bayes priors-with the latter being novel in the BSEM literature. Based on a simulation study, we find that these three default BSEM methods may perform very differently, especially with small samples. A careful prior sensitivity analysis is therefore needed when performing a default BSEM analysis. For this purpose, we provide a practical step-by-step guide for practitioners to conducting a prior sensitivity analysis in default BSEM. Our recommendations are illustrated using a well-known case study from the structural equation modeling literature, and all code for conducting the prior sensitivity analysis is available in the online supplemental materials.",
keywords = "Bayesian, CLASS SEPARATION, COMPONENTS, EMPIRICAL BAYES, FRAMEWORK, FREQUENTIST, INFERENCE, LINEAR MIXED MODELS, PARAMETERS, PRIOR DISTRIBUTIONS, SELECTION, default priors, sensitivity analysis, structural equation models",
author = "{van Erp}, S.J. and J. Mulder and Oberski, {Daniel L.}",
year = "2018",
doi = "10.17605/OSF.IO/5J3M9",
language = "English",
volume = "23",
pages = "363--388",
journal = "Psychological Methods",
issn = "1082-989X",
publisher = "AMER PSYCHOLOGICAL ASSOC",
number = "2",

}

Prior sensitivity analysis in default Bayesian structural equation modeling. / van Erp, S.J.; Mulder, J.; Oberski, Daniel L.

In: Psychological Methods, Vol. 23, No. 2, 2018, p. 363-388.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Prior sensitivity analysis in default Bayesian structural equation modeling

AU - van Erp, S.J.

AU - Mulder, J.

AU - Oberski, Daniel L.

PY - 2018

Y1 - 2018

N2 - Bayesian structural equation modeling (BSEM) has recently gained popularity because it enables researchers to fit complex models and solve some of the issues often encountered in classical maximum likelihood estimation, such as nonconvergence and inadmissible solutions. An important component of any Bayesian analysis is the prior distribution of the unknown model parameters. Often, researchers rely on default priors, which are constructed in an automatic fashion without requiring substantive prior information. However, the prior can have a serious influence on the estimation of the model parameters, which affects the mean squared error, bias, coverage rates, and quantiles of the estimates. In this article, we investigate the performance of three different default priors: noninformative improper priors, vague proper priors, and empirical Bayes priors-with the latter being novel in the BSEM literature. Based on a simulation study, we find that these three default BSEM methods may perform very differently, especially with small samples. A careful prior sensitivity analysis is therefore needed when performing a default BSEM analysis. For this purpose, we provide a practical step-by-step guide for practitioners to conducting a prior sensitivity analysis in default BSEM. Our recommendations are illustrated using a well-known case study from the structural equation modeling literature, and all code for conducting the prior sensitivity analysis is available in the online supplemental materials.

AB - Bayesian structural equation modeling (BSEM) has recently gained popularity because it enables researchers to fit complex models and solve some of the issues often encountered in classical maximum likelihood estimation, such as nonconvergence and inadmissible solutions. An important component of any Bayesian analysis is the prior distribution of the unknown model parameters. Often, researchers rely on default priors, which are constructed in an automatic fashion without requiring substantive prior information. However, the prior can have a serious influence on the estimation of the model parameters, which affects the mean squared error, bias, coverage rates, and quantiles of the estimates. In this article, we investigate the performance of three different default priors: noninformative improper priors, vague proper priors, and empirical Bayes priors-with the latter being novel in the BSEM literature. Based on a simulation study, we find that these three default BSEM methods may perform very differently, especially with small samples. A careful prior sensitivity analysis is therefore needed when performing a default BSEM analysis. For this purpose, we provide a practical step-by-step guide for practitioners to conducting a prior sensitivity analysis in default BSEM. Our recommendations are illustrated using a well-known case study from the structural equation modeling literature, and all code for conducting the prior sensitivity analysis is available in the online supplemental materials.

KW - Bayesian

KW - CLASS SEPARATION

KW - COMPONENTS

KW - EMPIRICAL BAYES

KW - FRAMEWORK

KW - FREQUENTIST

KW - INFERENCE

KW - LINEAR MIXED MODELS

KW - PARAMETERS

KW - PRIOR DISTRIBUTIONS

KW - SELECTION

KW - default priors

KW - sensitivity analysis

KW - structural equation models

U2 - 10.17605/OSF.IO/5J3M9

DO - 10.17605/OSF.IO/5J3M9

M3 - Article

VL - 23

SP - 363

EP - 388

JO - Psychological Methods

JF - Psychological Methods

SN - 1082-989X

IS - 2

ER -