Extending the mixed-effects model to consider within-subject variance for Ecological Momentary Assessment data

Area of Science:

• Statistics
• Psychometrics
• Behavioral Science

Background:

• Ecological Momentary Assessment (EMA) data collection yields rich, repeated measures of individual experiences.
• Traditional statistical models often oversimplify the complex within-subject (WS) variance inherent in EMA data.
• Modeling WS variance that varies with covariates and includes random subject effects is crucial for deeper insights.

Purpose of the Study:

• To present a novel statistical model for analyzing EMA data.
• To accommodate multiple random effects in both the mean and variance models.
• To allow WS variance to be influenced by covariates and random subject effects.

Main Methods:

• Development of a statistical model incorporating random intercepts and slopes for the mean.
• Inclusion of random effects for scale in the error variance model.
• Application of the model to a real-world EMA dataset and validation through simulation.

Main Results:

• The proposed model effectively captures complex WS variance structures in EMA data.
• Demonstrated significant benefits compared to simpler statistical approaches in simulation studies.
• The model provides a more nuanced understanding of individual variability and covariate influences.

Conclusions:

• The developed statistical model offers enhanced flexibility and accuracy for EMA data analysis.
• This approach is valuable for researchers seeking to model individual differences and time-varying effects.
• The findings highlight the importance of advanced statistical techniques for maximizing EMA data utility.