Estimating latent baseline-by-treatment interactions in statistical mediation analysis

Area of Science:

• Psychometrics
• Statistical Modeling
• Quantitative Psychology

Background:

• Statistical mediation analysis identifies mediators explaining treatment effects on outcomes.
• Moderation by baseline variables is common, but estimating baseline-by-treatment interaction (BTI) effects with latent variables is challenging.
• Existing methods lack guidance for latent BTI estimation in mediation.

Purpose of the Study:

• To provide and evaluate methods for estimating latent baseline-by-treatment interaction (BTI) effects in mediation analysis.
• To address the need for guidance on handling latent variables in BTI estimation.
• To offer practical solutions for researchers analyzing complex mediation models.

Main Methods:

• Discussed two general approaches: structural models and latent variable scoring prior to observed BTI estimation.
• Employed simulation studies to assess bias, power, type 1 error rates, and interval coverage.
• Illustrated methods with an applied example and provided R and Mplus syntax.

Main Results:

• Simulation results detailed the performance of the proposed methods for estimating latent BTIs and mediated effects.
• Evaluated bias, statistical power, Type 1 error rates, and confidence interval coverage.
• Demonstrated the practical utility of the methods through an applied example.

Conclusions:

• The study offers viable approaches for estimating latent BTI effects in mediation analysis.
• Provides researchers with tools and guidance for more accurate mediation analysis involving latent variables.
• Facilitates the implementation of advanced mediation models in psychological research.