A three-level mixed-effects location scale model with an application to ecological momentary assessment data.

Area of Science:

• Statistics
• Psychometrics
• Longitudinal Data Analysis

Background:

• Intensive longitudinal data collection, such as ecological momentary assessment (EMA), often requires analyzing within- and between-subject variance.
• Existing two-level mixed-effects models may not fully capture systematic day-to-day variations present in such data.

Purpose of the Study:

• To develop a flexible three-level mixed-effects location scale model for analyzing intensive longitudinal data.
• To allow covariates to influence variance at subject, day, and observation levels, improving upon existing two-level models.

Main Methods:

• Developed a three-level mixed-effects location scale model with a log-linear variance component.
• Utilized maximum likelihood estimation via SAS PROC NLMIXED for parameter estimation.
• Validated the model's accuracy through simulation studies and an adolescent mood EMA dataset.

Main Results:

• The proposed three-level model demonstrated significant benefits over the traditional two-level approach.
• Covariates were shown to influence variance at multiple levels (subject, day, observation) beyond their effects on means.
• The model effectively captured systematic day-to-day variations in variance.

Conclusions:

• The three-level mixed-effects location scale model provides a more comprehensive approach for analyzing complex longitudinal data structures.
• This method is valuable for studies with three-level data where joint modeling of mean and variance is of interest.
• The approach offers enhanced insights into within- and between-subject variability in intensive longitudinal studies.