A model for repeated clustered data with informative cluster sizes.

Area of Science:

• Statistics
• Psychiatry
• Epidemiology

Background:

• Chronic diseases often present with recurring episodes, where both episode frequency and severity indicate underlying condition severity.
• Traditional data collection at fixed intervals may not fully capture the nuances of recurring health events.
• Stressful life events and depression onset illustrate the need to analyze both frequency and intensity of events.

Purpose of the Study:

• To propose statistical models for analyzing data on both the frequency and severity of recurring events.
• To develop methods that account for latent severity influencing both event count and intensity.
• To provide a robust analytical framework for longitudinal health data.

Main Methods:

• Development of statistical models incorporating both event frequency and severity.
• Application of maximum likelihood estimators for parameter estimation.
• Simulation studies to evaluate estimator properties with small to moderate sample sizes.

Main Results:

• The proposed estimators demonstrate good finite sample properties.
• The statistical models are robust against model misspecification.
• The method was successfully applied to a real-world psychiatric dataset.

Conclusions:

• The developed models provide a powerful tool for analyzing complex longitudinal data where both event frequency and severity are informative.
• This approach enhances understanding of chronic conditions by better characterizing underlying latent severity.
• The findings have implications for research in mental health and other fields involving recurring health events.