Detection of Multiple Change Points in an Accelerated Failure Time Model Using Sequential Testing

• Activity Number: 79 - Contributed Poster Presentations: Lifetime Data Science Section
• Type: Contributed
• Date/Time: Monday, August 3, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Lifetime Data Science Section
• Author(s): Kristine Gierz* and Kayoung Park
• Companies: Old Dominion University and Old Dominion University
• Keywords: change point; sequential testing; accelerated failure time; survival analysis

Abstract

In survival analysis, the parametric accelerated failure time (AFT) model can be used as an alternative to the semi-parametric proportional hazards model. Whereas the proportional hazards model assumes that the hazard ratio is constant over time, the AFT model avoids this assumption but requires the assumption of distribution. The assumption that a hazard ratio is constant over time is often violated, and in practice there may be one or more change points in the hazard rate. There have been approaches suggested to identify multiple change points in a piecewise constant hazard model and in the proportional hazards model. We will propose a sequential testing approach to detect multiple change points in an AFT model with a Weibull distribution, since it is the only family of distributions to include the exponential distribution (with constant hazard) as a special case, and can be parameterized as a proportional hazards model. As in previous approaches to other models, the number of change points in the model need not be previously specified. Some numerical results based on simulated data and a real data example show that the method we propose can detect change points in the AFT model.