Robust and efficient estimation in the parametric proportional hazards model under random censoring

Area of Science:

• Statistics
• Biostatistics
• Survival Analysis

Background:

• The Cox proportional hazard regression model is a standard tool for analyzing censored lifetime data.
• While parametric Cox models offer efficiency, their maximum likelihood estimation is vulnerable to outliers.
• Robustness is crucial for reliable inference in real-world applications.

Purpose of the Study:

• To develop a robust estimation procedure for the parametric proportional hazards model.
• To address the limitations of maximum likelihood estimation in the presence of data contamination.
• To enhance the precision and reliability of statistical inference in survival analysis.

Main Methods:

• Development of a robust estimator using the minimum density power divergence approach.
• Theoretical justification of robustness through influence function analysis.
• Validation via simulations and real-world data examples.

Main Results:

• The proposed minimum density power divergence estimator demonstrates high robustness against data contamination.
• The estimator shows only a minor loss in efficiency with clean data.
• It provides more precise inference than likelihood-based methods and existing robust alternatives.

Conclusions:

• The minimum density power divergence approach offers a robust and efficient alternative for parametric proportional hazards modeling.
• This method enhances the practical applicability of proportional hazards models in fields like medicine and reliability.
• The findings support the use of this robust estimator for more dependable survival data analysis.