Kernel Smoothed Profile Likelihood Estimation in the Accelerated Failure Time Frailty Model for Clustered Survival...

Area of Science:

• Biostatistics
• Survival Analysis
• Biomedical Data Science

Background:

• Clustered survival data are common in biomedical research, particularly in family studies.
• Existing models may not fully capture the complexities of correlated event times within groups.
• Accelerated failure time (AFT) frailty models offer a flexible framework for such data.

Purpose of the Study:

• To develop a robust nonparametric maximum likelihood estimation (NPMLE) for AFT frailty models with clustered survival data.
• To address the statistical challenges posed by correlated event times within clusters.
• To provide a computationally feasible and statistically sound estimation method.

Main Methods:

• Development of a kernel smoother-aided Expectation-Maximization (EM) algorithm for NPMLE.
• Theoretical analysis establishing consistency, asymptotic normality, and semiparametric efficiency of the proposed estimator.
• Derivation of an EM-aided numerical differentiation method for variance estimation.

Main Results:

• The proposed NPMLE is shown to be consistent, asymptotically normal, and semiparametric efficient under appropriate bandwidth selection.
• The EM algorithm provides a practical computational approach for the complex estimation problem.
• Simulation studies confirm the good finite sample performance of the estimator.

Conclusions:

• The kernel smoother aided EM algorithm offers a powerful tool for analyzing clustered survival data using AFT frailty models.
• The method provides statistically reliable estimates for regression coefficients, crucial for identifying risk factors in clustered settings.
• Application to the Diabetic Retinopathy dataset demonstrates the practical utility of the approach in real-world biomedical studies.