Parametric and semiparametric estimation methods for survival data under a flexible class of models

Area of Science:

• Statistics
• Biostatistics
• Survival Analysis

Background:

• Accelerated failure time models in survival analysis typically assume linear covariate effects.
• This linear assumption is often too restrictive for real-world data.
• Flexible modeling of covariate effects is crucial for accurate survival analysis.

Purpose of the Study:

• To propose partially linear single index models for survival analysis.
• To incorporate flexible nonlinear relationships between covariates and transformed failure times.
• To develop and evaluate novel inference methods for these models.

Main Methods:

• Development of two inference methods: a weakly parametric global approximation and a semiparametric local quasi-likelihood approach.
• Establishment of asymptotic properties for both proposed inference methods.
• Application to a real-world example and extensive simulation studies.

Main Results:

• The proposed partially linear single index models effectively capture complex nonlinear covariate effects.
• Both developed inference methods demonstrate robust performance across various scenarios.
• Asymptotic properties of the methods are theoretically established.

Conclusions:

• Partially linear single index models offer a powerful and flexible extension to traditional accelerated failure time models.
• The proposed inference methods provide reliable tools for analyzing complex survival data with nonlinear covariate relationships.
• These advancements enhance the applicability and accuracy of survival analysis in practice.