Composite Quantile Regression For Censored Data

In This paper, we mainly studied the large sample properties for the local composite quantile regression estimator of regression function under censored data. Firstly, we introduce the basic ideas of quantile rgrssion, the mian results of the research and the forms of the censored data. Then we obtain the composite quantile regression of the censored data by applying loss estimation for weighted composite quantile regression. We take the high dimensional key density function and the K-M estimation of the censored data to be the weight of the loss function, what’s more, we take the censored data to estimate the value of non-parameter function at some point in this model, and we gain its asymptotic normality. From the result, we compare the asymptotic efficiency between the local composite quantile regression and the local least squares, analyze these relative efficiency, and verify the property of this method by using simulation study. We can get different results from different distribution of the error, so that we more fully prove the consequence. We main establish the model of the composite quantile regression for the censored data, because of incompletion for the data, we have to apply the property of the K-M estimation to simplify it in calculation, finally, we receive the more satisfactory results.