Title: Two-step estimation for varying coefficient regression models with censored data
Authors: Seong Jun Yang - Hankuk University of Foreign Studies (Korea, South) [presenting]
Cedric Heuchenne - University of Liege (Belgium)
Ingrid Van Keilegom - KU Leuven (Belgium)
Abstract: Estimators of the coefficient functions for the varying coefficient model are proposed where the response is subject to random right censoring. The model includes different coefficient functions depending on various covariates. Since multivariate smoothing is unavoidable under the model, smooth backfitting is applied to avoid ``the curse of dimensionality". The estimation method is based on the Bucklely-James type transformation, where the estimators achieved by Koul-Susarla-Van Ryzin type transformation are used for primary estimators of the coefficient functions. Asymptotic normality of the proposed estimators are given, and numerical studies are shown to illustrate the reliability of the estimators.