Strong Consistency of Non-parametric Regression Estimates with Censored Data

Let (X,Y) be an R~d×R valued random vector with E|Y|<∞ and(X_1,Y_1) (X_2,Y_2), …, (X_n,Y_n) be i.i.d.observations of (X,Y). To estimate the regression function m(x)=E(Y|X=x), Stone suggested m_n(x)=sum from i=1 to n(W_(ni)(x)Y_i), where W_(ni)(x)=W_(ni)(x,X_1,X_2,…,X_n)(i=1,2,…,n) are weight functions. Devroye and Chen Xiru established the strong consistency of m_n(x). In this paper, we discuss the case that{Y_i} are censored by {t_i}, where{t_i} are i.i.d. random variables and also independent of{Y_i}. Under certainconditions we still obtain the strong consistency of m_n(x).