Building a simpler moderated nonlinear factor analysis model with Markov Chain Monte Carlo estimation

Area of Science:

• Psychometrics
• Statistical Modeling

Background:

• Moderated nonlinear factor analysis (MNLFA) is a key tool in psychometric research and integrative data analysis.
• It unifies several modeling traditions and extends them by linking latent variable heterogeneity to differential item functioning.

Purpose of the Study:

• To demonstrate a flexible Markov chain Monte Carlo (MCMC)-based approach for MNLFA.
• To highlight statistical and practical advantages over traditional likelihood-based estimation.

Main Methods:

• Utilized a Markov chain Monte Carlo (MCMC) approach for MNLFA.
• Implemented enhancements including missing data handling, multiply imputed factor scores, diverse data type support, residual diagnostics, and manifest-by-latent variable interactions.
• Integrated with regression modeling strategies and graphical diagnostics.

Main Results:

• The MCMC approach provides statistical and practical enhancements for MNLFA.
• The method effectively handles incomplete moderators and various data types (continuous, binary, ordinal, count).
• Novel diagnostics and interaction effects facilitate robust analysis and interpretation.

Conclusions:

• The illustrated MCMC approach offers a powerful and flexible alternative for MNLFA.
• This method enhances psychometric analyses by improving data handling and diagnostic capabilities.
• The approach integrates seamlessly with existing statistical practices and software.