A hybrid hazard-based model using two-piece distributions.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Cox proportional hazards models are standard for survival analysis but assume constant relative hazards.
• Predicting specific survival quantiles, like median survival time, benefits from flexible parametric baseline distributions.
• Existing models may lack flexibility in capturing diverse hazard function shapes.

Purpose of the Study:

• To propose novel, flexible hazard-based models for right-censored survival data.
• To incorporate a wide range of two-piece asymmetric baseline distributions for enhanced model adaptability.
• To characterize covariate effects using time-scale changes in hazard progression and relative hazard ratios.

Main Methods:

• Development of flexible hazard-based models utilizing two-piece asymmetric baseline distributions.
• Implementation of parametric, semi-parametric (partly linear), and non-parametric covariate effect forms.
• Application of full likelihood and profile (local) likelihood estimation techniques for parameter estimation.

Main Results:

• The proposed models offer flexibility in accommodating various hazard function shapes.
• Covariate effects are modeled through time-scale modifications, allowing for diverse functional forms.
• Simulation studies demonstrate the finite-sample performance of the developed methods.

Conclusions:

• Flexible hazard-based models with asymmetric distributions provide a powerful alternative for survival data analysis.
• The proposed methods enhance the prediction of survival outcomes and the understanding of covariate effects.
• The approach is validated through simulations and illustrated with real-world data applications.