Expectile Regression for Censored Data Based on Data Augmentation.

Area of Science:

• Statistics
• Biostatistics
• Survival Analysis

Background:

• Statistical modeling of censored survival data is crucial in biomedical applications.
• Existing methods like censored quantile and expectile regression have limitations, including computational challenges and reliance on estimating unknown survival functions.

Purpose of the Study:

• To develop a novel and unified estimation method for expectile regression with heterogeneous censored survival data.
• To address limitations of existing methods by employing a data augmentation approach instead of inverse probability of censoring weighting (IPW).

Main Methods:

• Investigated expectile regression for censored data to capture heterogeneity.
• Developed a unified estimation method using data augmentation, avoiding estimation of the survival function for censored times.
• Evaluated the proposed approach through extensive simulation studies and analysis of two real datasets.

Main Results:

• The proposed data augmentation method effectively handles various censoring mechanisms for expectile regression.
• Simulation studies demonstrated the performance of the novel approach.
• Analysis of real datasets yielded intriguing findings, highlighting the method's practical utility.

Conclusions:

• The developed method offers a computationally feasible and unified approach for analyzing heterogeneous censored survival data.
• The R function DAer implements the proposed algorithm, facilitating its application in bio-medicine.
• The findings underscore the value of expectile regression with data augmentation for survival data analysis.