An Improved Estimation Using Polya-Gamma Augmentation for Bayesian Structural Equation Models with Dichotomous Variables. Issue 2 (3rd April 2018)

• Record Type: Journal Article
• Authors: Kim, Seohyun; Lu, Zhenqiu; Cohen, Allan S.
• Is Part Of: Measurement. Volume 16:Issue 2(2018)
• Page Start: 81
• Page End: 91
• Publication Date: 2018-04-03
• Subjects: Structural equation modeling -- Bayesian algorithms -- Polya-Gamma distribution; Social sciences -- Methodology -- Periodicals; Social sciences -- Statistical methods -- Periodicals; Social sciences -- Mathematical models -- Periodicals; Measurement -- Periodicals; 300.72
• DOI: 10.1080/15366367.2018.1437303
• Languages: English
• ISSNs: 1536-6367

Abstract

Abstract: Bayesian algorithms have been used successfully in the social and behavioral sciences to analyze dichotomous data particularly with complex structural equation models. In this study, we investigate the use of the Polya-Gamma data augmentation method with Gibbs sampling to improve estimation of structural equation models with dichotomous variables. An empirical example is provided to illustrate the performance of different estimation approaches followed by a simulation study to evaluate the proposed method. The Polya-Gamma method is shown to provide stable results with larger effective sample size than standard Gibbs sampling.