A latent variable approach to jointly modeling longitudinal and cumulative event data using a weighted two-stage...

Area of Science:

• Psychology
• Biostatistics
• Health Informatics

Background:

• Ecological momentary assessment (EMA) collects real-time data on psychological, behavioral, and contextual states in mHealth studies.
• EMA data aids in understanding temporal dynamics and relationships between states and adverse health events.
• Analyzing longitudinal EMA data is crucial for understanding complex health behaviors like smoking.

Purpose of the Study:

• To propose a joint statistical model for analyzing longitudinal EMA data.
• To determine the association between latent psychological states and repeated cigarette use.
• To address partially unobservable predictors and outcomes in smoking cessation studies.

Main Methods:

• A dynamic factor model for longitudinal submodel to track time-varying latent states.
• A Poisson regression model for cumulative risk submodel connecting latent states to event counts.
• A two-stage estimation approach using importance sampling-based weights to mitigate bias.

Main Results:

• The proposed joint model effectively analyzes longitudinal EMA data with unobservable components.
• Importance sampling weights successfully reduced bias in cumulative risk submodel parameters.
• Above-average negative mood intensities were significantly associated with increased cigarette smoking.

Conclusions:

• The developed joint model provides a robust method for analyzing complex EMA data in health studies.
• Psychological states, particularly negative mood, play a significant role in repeated cigarette use.
• This research offers insights into addiction mechanisms and informs smoking cessation interventions.