Checking Methods for Accelerated Failure Time Models

• Activity Number: 367 - Contributed Poster Presentations: ENAR
• Type: Contributed
• Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
• Sponsor: ENAR
• Author(s): D Choi* and Woo Jung Bae and S Kang
• Companies: Duke University and University of Florida and Yonsei University
• Keywords: AFT models; Omnibus test; Link function test; Functional form test; Gaussian process; Kolmogorov type supremum test

Abstract

Since the direct interpretation of failure time is convenient, an Accelerated Failure Time (AFT) models provide nice alternatives to the popular Cox models. In this article, the research is conducted to verify whether the observed data fit the AFT models by checking an omnibus, a link function, and a functional form test based on the martingale residuals. The Kolmogorov type supremum test is used to check the model assumptions of AFT models objectively. Numerical approaches are considered for testing the suitability of the semi-parametric univariate AFT models. In addition, test statistics and their consistencies are provided by showing that they follow the zero-mean Gaussian processes.