Capturing Heterogeneity in Levels, Variability, and Couplings across Persons and Time with a Hierarchical...

Area of Science:

• Psychological Science
• Statistical Modeling
• Longitudinal Data Analysis

Background:

• Conventional time-varying coefficient modeling (TVCM) primarily examines directional effects.
• Existing methods are limited in exploring undirected associations (couplings) and variability over time.
• Intensive longitudinal data offers rich insights into dynamic psychological processes.

Purpose of the Study:

• To extend TVCM by incorporating a multivariate normal distribution formulation.
• To enable the analysis of changing undirected associations and variability.
• To provide hierarchical models capturing interindividual differences in dynamic change patterns.

Main Methods:

• Developed aggregate-level and two hierarchical TVCM versions.
• Utilized person-specific intercepts for onset differences and partial pooling for coefficient functions.
• Applied models to intensive longitudinal data from anxiety patients undergoing therapy.

Main Results:

• Demonstrated the ability to model unfolding changes in levels, volatility, and couplings of nervousness and threat monitoring.
• Quantified between-person heterogeneity in these dynamic processes.
• Showcased how coefficient function derivatives identify periods of stability and change.

Conclusions:

• The extended TVCM framework offers a powerful tool for psychological research on dynamic processes.
• The proposed models capture complex interindividual differences in change trajectories.
• Future extensions can incorporate person-level characteristics to explain heterogeneity and predict outcomes.