Factored structural equation modeling in blimp

Area of Science:

• Behavioral and social sciences
• Statistical modeling
• Psychometrics

Background:

• Multivariate structural equation modeling (SEM) can be complex to implement.
• Existing SEM approaches may struggle with diverse data types and structures.
• Factored structural equation modeling (FSEM) presents a novel alternative.

Purpose of the Study:

• Introduce factored structural equation modeling (FSEM).
• Demonstrate FSEM implementation using Blimp software and the rblimp R package.
• Provide a flexible framework for complex data analysis in social sciences.

Main Methods:

• Reconceptualizes joint distributions using univariate/multivariate submodels.
• Specifies models via regression equations.
• Treats latent variables as missing data imputed via Bayesian data augmentation.
• Accommodates continuous, binary, ordinal, nominal, count, and two-part variables.
• Handles interactions, nonlinear effects, heteroscedasticity, and multilevel data.

Main Results:

• FSEM seamlessly integrates diverse variable types and complex data structures.
• The approach avoids violating distributional assumptions.
• Illustrative models range from confirmatory factor analysis to dynamic, multilevel, and hybrid generalized-linear-structural equation models.
• Provides Blimp syntax and real-data examples.

Conclusions:

• FSEM offers a user-friendly and flexible framework for advanced statistical modeling.
• The method is suitable for researchers in behavioral and social sciences.
• Future research directions and limitations are discussed.