A shared frailty regression model for clustered survival data.

Area of Science:

• Biostatistics
• Survival Analysis
• Reliability Engineering

Background:

• Frailty models are crucial for analyzing correlated survival data in medical research.
• Existing models often require arbitrary choices for weighting mixture components.
• Quantifying dependence in multivariate survival data remains a challenge.

Purpose of the Study:

• To propose a new multivariate frailty model utilizing a mixture of inverse Gaussian distributions.
• To enhance dependence quantification and simplify parameter estimation in frailty modeling.
• To demonstrate the model's efficacy and advantages over existing approaches.

Main Methods:

• Development of a novel frailty model based on a mixture of inverse Gaussian distributions.
• Utilizing the expectation-maximization algorithm for parameter estimation via hierarchical representation.
• Derivation of a closed-form Laplace transform for Kendall's tau calculation.
• Numerical assessment via Monte Carlo simulations and application to cancer datasets.

Main Results:

• The proposed frailty model offers direct weight parameterization, eliminating arbitrary choices.
• The closed-form Laplace transform enables straightforward quantification of the Kendall's tau measure of dependence.
• Expectation-maximization provides a more stable and simpler estimation method.
• Simulations and real-world data analysis confirm the model's benefits over existing frailty models.

Conclusions:

• The novel inverse Gaussian mixture frailty model provides a statistically sound and practical advancement in multivariate survival analysis.
• The model facilitates improved understanding of dependence structures and parameter estimation.
• The methodology is implemented in the R package extrafrail for broader accessibility.