In recent years , a lot of econometric literatures have been devoted to non-parametric estimation in semiparametric models , there are profile-kernel method and backfitting method as so on smoothing. This paper considers smoothing spline method with longitudinal data , supposes that there are n subjects and for the i-th subject, there are n1 observations. The model is Yij = XijT β + g(ti) +εi(tij), where (Yij,Xij,tij) are observations , β = (β0, … ,βk)T,g{t) is arbitrary smooth function of t , and the error term εi(t) is a zero-mean , identically stochastic process. The observations are assumed to be independent for different subjects , β and g(t) are needs to be estimated.This paper presents the spline estimation and algrithm for every individual which has the same or different observation . And then we consider the nonparamet-ric function g(t) , when it's a monotone increasing function , we get the expression of the estimation . For the smoothing parametric , we give a cross-validation rule.
|