Non-normal GMM with Covariates: A Modified 3-Step Analysis
Journal
Structural Equation Modeling: A Multidisciplinary Journal
Journal Volume
32
Journal Issue
4
Start Page
606-617
ISSN
1070-5511
1532-8007
Date Issued
2025-04
Author(s)
Abstract
The multi-step bias correction analysis of GMM with covariates typically assumes a normal distribution in the first step. Thereafter, the analysis undertakes the control of classification errors and the incorporation of further covariates for the purpose of further analysis. However, it is important to note that assuming normality in the presence of nonnormal GMM can lead to biased corrections. To achieve unbiased GMM bias correction analysis, this study proposes a modified 3-step analysis. Through simulation studies, this research demonstrates the bias arising from ignoring nonnormal distributions in GMM bias correction analysis. Empirical examples are provided to illustrate the construction steps of the modified 3-step analysis, along with detailed Mplus syntax.
Subjects
BCH
covariates
GMM
modify 3-step analysis
non-normal
Publisher
Routledge
Type
journal article