Discrete-Time Survival Factor Mixture Analysis for Low-Frequency Recurrent Event Histories

Area of Science:

• Statistics
• Survival Analysis
• Latent Variable Modeling

Background:

• Traditional survival models often struggle with recurrent events and unobserved heterogeneity.
• Analyzing discrete-time survival data with multiple events per subject requires specialized statistical frameworks.

Purpose of the Study:

• To extend latent class analysis for discrete-time recurrent event data.
• To introduce a partial gap time model for analyzing low-frequency recurrent event histories.
• To account for event-specific hazards, recurrences within periods, and correlations in event times.

Main Methods:

• Latent class analysis (LCA) framework extension.
• Development of a partial gap time model, parameterized as a restricted factor mixture model.
• Application to juvenile offending data to illustrate model capabilities.

Main Results:

• The proposed model successfully accommodates event-specific baseline hazard probabilities and covariate effects.
• The model handles event recurrences within a single time period.
• It accounts for both within- and between-subject correlations of event times.

Conclusions:

• The developed model expands the family of latent variable survival models.
• Researchers can now explicitly address unobserved heterogeneity in event timing across the lifespan.
• This provides a more nuanced understanding of recurrent event processes.