A Bivariate Mixed-Effects Location-Scale Model with application to Ecological Momentary Assessment (EMA) data

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Mixed-Effects Models

Background:

• Analyzing multiple, correlated outcomes over time requires sophisticated statistical methods.
• Traditional models often assume homogeneity of variance, which may not hold in real-world data.
• Understanding both within-subject (WS) and between-subject (BS) variability is crucial for accurate interpretation.

Purpose of the Study:

• To propose a bivariate mixed-effects location-scale model for joint analysis of two continuous outcomes.
• To investigate the within-subject (WS) and between-subject (BS) associations between outcomes and their covariate relationships.
• To model heterogeneous variances and incorporate random scale effects for enhanced flexibility.

Main Methods:

• Developed a bivariate mixed-effects location-scale model.
• Jointly modeled two continuous outcomes (positive affect, boredom) from repeated measures.
• Incorporated covariates to explain heterogeneity in BS and WS variances and associations.
• Extended WS variance models with random scale effects.

Main Results:

• The WS association between positive affect and boredom was negative and significantly related to covariates.
• Both BS and WS variances were heterogeneous for the studied outcomes.
• The variance of the random scale effects was statistically significant, confirming model utility.

Conclusions:

• The proposed location-scale model effectively captures complex associations and heterogeneity in longitudinal data.
• Adolescent mood states exhibit significant WS associations influenced by covariates.
• The model's flexibility in handling heterogeneous variances and random scale effects improves the analysis of repeated measures.