论文部分内容阅读
When nonparametric regression models arc used on longitudinal or cluster data,there have been active studies on how the asymptotic e?ciency of a nonparametric function estimator depends on the handling of the within-cluster correlation.In particular, the smoothing splines and the local polynomial kernels exhibit di?erent behaviors.In this paper, we show that the generalized estimation equations based on weighted least squares regression splines for the nonparametric function have a rather interest ing property: the asymptotic bias of the estimator does not depend on the working correlation matrix, but the asymptotic variance, and therefore the mean squared error, is minimized when the true correlation structure is speci?ed.This property on the asymptotic bias distinguishes the regression splines from the smoothing splines.