Counterfactual Mediation Analysis with a Latent Class Exposure

Area of Science:

• Developmental Psychology
• Biostatistics
• Epidemiology

Background:

• Latent classes are valuable in developmental research for identifying distinct subgroups.
• Integrating latent classes into counterfactual mediation models presents significant methodological challenges.
• Existing methods for relating latent class exposure to distal outcomes have limitations in bias and flexibility.

Purpose of the Study:

• To develop and validate a novel method, updated pseudo class draws (uPCD), for analyzing latent class exposure and distal outcomes within counterfactual mediation frameworks.
• To assess the performance of uPCD compared to existing techniques using simulated data.
• To provide a flexible approach for incorporating latent groupings into various counterfactual mediation methods.

Main Methods:

• Simulated data based on the Avon Longitudinal Study of Parents and Children.
• Development of the updated pseudo class draws (uPCD) method, extending pseudo class draws by using multiply imputed class membership probabilities.
• Comparison of uPCD against established methods: one-step, bias-adjusted three-step, modal class assignment, non-inclusive pseudo class draws, and inclusive pseudo class draws.

Main Results:

• The uPCD method demonstrated minimal bias in estimating counterfactual mediation effects across all entropy levels.
• uPCD performance was comparable to recommended methods like one-step and bias-adjusted three-step.
• uPCD offers enhanced flexibility for integrating latent class variables into diverse counterfactual mediation models.

Conclusions:

• Updated pseudo class draws (uPCD) provide a robust and flexible solution for analyzing latent class exposures in counterfactual mediation models.
• The uPCD method effectively addresses bias issues, offering a reliable tool for developmental researchers.
• This new approach expands the scope of latent class analysis within complex mediation frameworks.