A Two-Step Robust Estimation Approach for Inferring Within-Person Relations in Longitudinal Design: Tutorial and...

Area of Science:

• Psychological research methods
• Longitudinal data analysis
• Causal inference

Background:

• Disaggregating within-person variability from between-person differences is a key challenge in psychological research.
• Existing methods may struggle with complex relationships and unobserved confounders.

Purpose of the Study:

• To introduce and demonstrate a new two-step approach for disaggregating within-person variability.
• To provide a tutorial, simulation, and example of the proposed method.
• To enhance causal parameter estimation in psychological research.

Main Methods:

• A two-step procedure involving structural equation modeling to predict within-person variability scores (WPVS).
• Estimation of causal parameters using a potential outcome approach, specifically structural nested mean models (SNMMs).
• Investigation of estimation performance through large-scale simulations with longitudinal data (T>=4).

Main Results:

• The proposed method allows flexible inclusion of curvilinear and interaction effects for WPVS.
• It offers more accurate causal parameter estimates for reciprocal relations, even with unobserved confounders.
• The approach reduces the risk of improper solutions and does not require models for time-varying confounders.

Conclusions:

• The new approach effectively disaggregates within-person variability and improves causal inference.
• It performs well under various conditions with sufficient longitudinal data.
• The method is illustrated with an example from the Tokyo Teen Cohort (TTC) study.