Title: A general class of semiparametric accelerated rate models for recurrent event processes under informative censoring
Authors: Sy Han Chiou - University of Texas at Dallas (United States) [presenting]
Abstract: Recurrent event analyses in practice often face two challenges: existing model formulations for covariate effects do not fit the data well, and the censoring time is informative after conditioning on covariates. We tackle both challenges in a general class of semiparametric models, which includes proportional rate model, the accelerated rate model, the semiparametric transformation model, and the scale-change model as special cases. Informative censoring is allowed for through a subject-level frailty whose distribution is left unspecified beyond unit expectation. The model parameters are estimated in a two-step procedure. The asymptotic properties of the resulting estimator are established, with the asymptotic variance estimated from a novel resampling strategy. The structure of the proposed model enables model specification tests for each subclass of models, an important issue which has not been well studied. Numerical studies demonstrate that the proposed estimator and tests have attractive performance under both noninformative and informative censoring.