CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Transformed dynamic quantile regression on censored data Authors:  Tony Sit - The Chinese University of Hong Kong (Hong Kong) [presenting]
Abstract: A class of power-transformed linear quantile regression models is proposed for time- to-event observations that are subject to conditionally independent censoring. By introducing power transformation with different transformation parameters for individual quantile levels, the proposed model relaxes the limitation due to the global linear assumption required in a previous formulation, and thus, it enjoys a greater extent of flexibility. Our framework provides simultaneous estimation of various quantiles and is amongst the first which considers power transformation as a process with respect to the quantile values. The uniform consistency and weak convergence of the proposed estimator as a process with respect to a sequence of quantile levels are established. Simulation studies and applications to real data sets are presented for verifying the performance of the proposed estimator our empirical results is shown to outperform existing contenders under various settings.