Comparing Censored Quantile Regression Models in Prediction Performances

• Activity Number: 283
• Type: Invited
• Date/Time: Tuesday, August 11, 2015 : 8:30 AM to 10:20 AM
• Sponsor: IMS
• Author(s): Ruosha Li* and Limin Peng
• Companies: The University of Texas School of Public Health and Emory University
• Keywords: Censored quantile regression ; Misspecification ; Model comparisons ; Prediction performance ; Perturbation resampling ; Survival analysis

Abstract

Quantile regression is increasingly recognized as a powerful tool for estimating and predicting the quantiles of a survival outcome. The predicted quantile survival times are easy to interpret and facilitate comprehensive insights into the disease progressions. However, there lacks a rigorous method for evaluating and comparing censored quantile regression models in terms of prediction performances. The article proposes a sensible framework to bridge this gap. The perturbation-based inferential procedures are robust to the realistic complication of model mis-specification, which greatly enhances the practical utilities of the proposed methods. Extensive simulations and a real data example demonstrate satisfactory performances of the proposed methods.