Bayesian Analysis of Multigroup Nonlinear Structural Equation Models With Polytomous Data Apply to Educational Research

• In Event: AERA Poster Session 7
  In Poster Session: Division D Section 2 Poster Session 2

Abstract

A Bayesian approach for analyzing multi-group nonlinear SEMs with ordered categorical variables was proposed to investigate the relationships among all identified influential factors that have an impact on the student interest in science and different patterns of motivations for the student interest concerning different groups of students by mixing the multiple methodologies. The discrete variables were handled by the threshold approach to avoid the estimation errors produced by the skewness of discrete variables collected through the questionnaires. Data augmentation strategies based on the Gibbs sampler were used to estimate parameters for situations with small sample sizes, and the deviance information criterion (DIC) values help us select the most accurate model associated with the constraints across the groups.

Authors

• Jihang Chen, Boston College

• Zhushan Mandy Li, Boston College