Multivariate survival analysis with positive stable frailties

Area of Science:

• Biostatistics
• Statistical Modeling
• Survival Analysis

Background:

• Dependent multivariate survival data presents analytical challenges.
• Existing models may not adequately capture the complexities of correlated survival outcomes.
• The need for flexible and robust statistical frameworks is crucial in biostatistical research.

Purpose of the Study:

• To develop and describe a Bayesian modeling approach for dependent multivariate survival data.
• To incorporate positive stable frailty distributions for enhanced modeling flexibility.
• To illustrate the methodology's application using real-world kidney infection data.

Main Methods:

• Bayesian modeling framework for multivariate survival data.
• Utilized positive stable frailty distributions to model dependence.
• Employed a piecewise exponential model with a correlated prior for baseline hazard.
• Implemented a modified Gibbs sampling procedure for parameter estimation.

Main Results:

• Successfully estimated the stable law parameter and conditional proportional hazards model parameters.
• Demonstrated the practical application and efficacy of the proposed Bayesian methodology.
• The modified Gibbs sampling facilitated efficient estimation in complex survival data scenarios.

Conclusions:

• The proposed Bayesian approach offers a flexible and effective method for analyzing dependent multivariate survival data.
• Positive stable frailty distributions provide a valuable tool for capturing dependence structures.
• The methodology is well-suited for applications in biostatistics, such as the analysis of infectious disease data.