Estimation of Varying Coeffcient Model with Randomly Right-Censored Covariate

Data is often right-censored due to loss of follow-up, drop-out from experiments or end of clinical trials. The treatment of right-censored data has attracted the interest of many researchers. Most existing studies focus on the cases where the response variable is censored. Predictors in the regression model may also suffer from right-censoring. However, there are only sporadic works on the treatment of censored covariates. In this paper, the estimation of a varying coeffcient model with randomly rightcensored covariates is investigated. To deal with the right-censoring, we adjust the objective function directly through an inverse probability weighting, instead of adjusting the right-censored predictor. Estimation of the regression coeffcient is proposed. The asymptotic properties of the proposed estimator are rigorously investigated. Numerical simulations and real-data analyses demonstrate that the proposed method has good finite sample properties.