CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Generalized partially linear single-index cure mixture models with interval-censored data Authors:  Xiaoyu Liu - Nanyang Technological University (Singapore) [presenting]
Liming Xiang - Nanyang Technological University (Singapore)
Abstract: The mixture cure model, typically combined the Cox proportional hazards model as the latency component for event time and logistic regression as the incidence component for the probability of cure, is often used to analyse survival data from subjects when a subset of them will never experience the event of interest. However, it is not realistic in some practical cases to assume the cure probability as a known transformation of a linear combination of covariates. We propose a double semiparametric mixture cure model for interval-censored data, allowing nonlinear effects of covariates on the cure probability through a generalized partially single-index model. We develop a Bayesian inference procedure for estimation based on a two-stage data augmentation method for deal with interval censored data, and polynomial splines for approximating nonlinear functions in both components of the proposed model. Simulation results demonstrate the finite sample performance of the proposed Bayesian procedure. To illustrate the proposed method, we apply the proposed procedure to analyse the data from a hypobaric decompression sickness study.