Penalized estimation of general frailty Poisson models for recurrent count events

Area of Science:

• Biostatistics
• Statistical modeling
• Survival analysis

Background:

• Panel count data present unique challenges in statistical analysis.
• Frailty models are essential for handling correlated event times in clustered data.
• Efficient estimation methods are crucial for complex models like frailty models.

Purpose of the Study:

• To develop spline-based efficient estimation methods for frailty models with panel count data.
• To propose a computationally efficient algorithm for analyzing such data.
• To introduce a flexible estimation approach and a score test for overdispersion.

Main Methods:

• Spline-based penalization techniques for efficient estimation.
• A two-stage iterative expectation-maximization algorithm.
• General quasi-likelihood estimation for flexible model fitting.
• A score test for detecting overdispersion in count data.

Main Results:

• The proposed methods provide efficient and flexible estimation for frailty models.
• The developed algorithm is computationally efficient and easy to implement.
• The score test effectively detects overdispersion in panel count data.
• The methods are validated through extensive simulations and a real-world study.

Conclusions:

• Spline-based frailty models offer a powerful approach for analyzing panel count data.
• The proposed estimation and testing procedures are statistically sound and practically useful.
• These methods enhance the analysis of correlated count data, particularly in biomedical research.