摘 要: | Let (X, Y), (X_1, Y_1), …, (X_n, Y_n) be i. i. d. random vectors taking values in R_d×Rwith E(|Y|)<∞. To estimate the regression function m (x) = E (Y|X= x), we use thekernel estimate m_n(x)= sum from i=1 to n K((X_j-x)/h_n) where K(x) is a kernel functionand h_n a window width. In this paper, we establish the strong consistency of m_n(x) whenE(|Y|~P)<∞ for some p>l or E{exp(t|Y|~λ)}<∞ for some λ>0 and t>O. It is remakablethat other conditions imposed here are independent of the distribution of (X, Y).
|