Blanche variable is nothing but a conditional

Blanche et al. (2013) modi ed the IPCW method and proposed the method called conditional inverse probability of censoring weighting (CIPCW). In this method, the observations of uncensored individuals at time t are weighted by the conditional probability of being uncensored given the marker, instead of weighting by the marginal probability. The main advantage of this method over IPCW method is that it does not assume the independence between a marker and censoring time. Hung and Chiang (2011) proposed a non-parametric estimation method for the time-dependent partial area under the ROC curve. Recently, Li et al. (2016) extended the nearest neighborestimation method and proposed a non-parametric method called the Simple Method. This method introduces a weight variable and defi ne the sensitivity and specfi city estimate based on this weight variable. The weight variable is nothing but a conditional survival function of T given the observeddata. This method use the available data very efficiently as compared to all the other methods. The advantage of this method over the nearest neighbor estimation method is that it is robust to the choice of bandwidth (Li et al., 2016). For a comprehensive summary of the time-dependent ROC curve estimation methods except for the simple method, the reader is referred to Blanche et al. (2013) and Kamarudin et al. (2017).The ROC curve estimation methods presented above are developed under the classical survival method assumptions. In survival analysis, we typicallyneed the non-informative censoring assumption and correctly specifying a regression model, e.g. the proportional hazards model. Schmid and Potapov (2012) conducted simulations to study the e ects of model misspecfi cation and censoring rate on the behaviors of the classi cation accuracy measuresestimators. From the simulation results, they found that censoring rates and model misspecfi cation have signi cant e ects on the classi cation accuracy measures estimators. Schmid et al. (2013) also conducted similar simulation study to assess the effect of model misspecfi cation on medical decision making. From the simulations results, they found that the ICPW method is insensitive to the model assumptions violation as compared to the other considered methods and hence they recommended this method to evaluate the classi cation accuracy of survival models.