Comparing multiple statistical software for multiple-indicator, multiple-cause modeling: an application of gender...

Area of Science:

• Psychometrics
• Statistical Modeling
• Cognitive Neuroscience

Background:

• The multiple-indicator, multiple-cause (MIMIC) model is a specialized form of structural equation modeling (SEM) within a latent variable framework.
• It integrates covariates of interest into factor analysis, offering rigorous results and broad availability in statistical software.

Purpose of the Study:

• To introduce the multiple-indicator, multiple-cause (MIMIC) model.
• To demonstrate its implementation using SAS CALIS, R lavaan, and Mplus.
• To examine gender disparities in cognitive functioning using the MIMIC model.

Main Methods:

• The study detailed the formulation, specification, and identification of the MIMIC model.
• Empirical application involved the Midlife in the United States II (MIDUS II) Study dataset (N=4109).
• Analyses were conducted using SAS CALIS procedure, R lavaan package, and Mplus version 8.0, presenting syntaxes for input, output, and diagrams.

Main Results:

• All three software packages could utilize raw data and empirical covariance matrices.
• Mplus offers greater modeling flexibility but limited data manipulation compared to SAS and R.
• Results indicated men outperform women in executive function, while women exhibit superior episodic memory.

Conclusions:

• The study validated the utility of the MIMIC model with three popular statistical software packages.
• Findings align with previous empirical research on gender differences in cognitive functioning.
• Coding procedures and examples are provided as a tutorial for researchers interested in latent construct modeling.