Title: Accelerated failure time models with logconcave errors
Authors: Ruixuan Liu - Emory University (United States) [presenting]
Abstract: Accelerated failure time models are studied where the survivor function of the error term is logconcave. The logconcavity assumption is natural to represent the ageing or wear-out effect of the failure time and it is often assumed in related economic models. We construct semiparametric maximum likelihood estimates subject to the shape restriction with right censored data. We establish large sample properties of the resulting estimates for the finite dimensional parameter and show our estimators achieve the semiparametric efficiency bound. Our shaped restricted estimation also facilitates the computation, as in the proposed iterative procedure it leads to monotone estimating equations in each iteration step.