A latent variable mixed-effects location scale model that also considers between-person differences in the...

Area of Science:

• Public health research
• Psychology
• Statistics

Background:

• Intensive longitudinal data from experience sampling and daily diary designs are increasingly common in public health.
• Mixed-effects models and mixed-effects location scale models are typically used for analyzing this data.
• Existing models may not fully account for measurement error or individual variability in dynamic processes.

Purpose of the Study:

• To introduce an extension of the mixed-effects location scale model that incorporates measurement error via a latent factor model.
• To allow for person-specific differences in residual variance and autoregressive processes of the latent factor.
• To provide a maximum likelihood estimation method and compare its performance with a Bayesian approach.

Main Methods:

• Development of an extended mixed-effects location scale model incorporating a latent factor model for measurement error.
• Estimation of model parameters using a maximum likelihood approach.
• Comparison of maximum likelihood with a Bayesian estimation approach through a simulation study.

Main Results:

• The proposed extended mixed-effects location scale model effectively accounts for measurement error.
• The model allows for significant between-person variability in latent factor residual variance and autoregressive parameters.
• Simulation results indicate good performance of the maximum likelihood estimation approach.

Conclusions:

• The enhanced mixed-effects location scale model offers a powerful tool for analyzing intensive longitudinal data with measurement error.
• The model provides a more nuanced understanding of individual differences in dynamic psychological processes.
• Future research can extend this framework to address more complex longitudinal data structures and research questions.