Séance Séminaire

Séminaire des Doctorant·e·s

Estimation non-paramétrique pour un modèle de censure.

In censorship models the random variable of interest is subject to random censoring by another random variable. In this framework, we build a new kernel estimator based on the so-called synthetic data of the mean squared relative error for the regression function. We establish its strong uniform convergence with a rate over a compact set and asymptotic normality. The asymptotic variance is explicitly given and as a product we give confidence bands. A simulation study is conducted to support our theoretical results and show the advantages compared with other methods. Finally, we apply our methodology to real data to show the superiority of the relative error regression (RER) estimator to other estimators.