Some topics on robust nonparametric regression and regression quantile

Local kernel estimates and B-spline estimates are considered in the nonparametric regression and the generalized linear models. The consistency and asymptotic normality of kernel estimates are proved. Simulations on B-spline estimates for nonparametric regression and generalized linear models are provided.;Also a sequential procedure based on the regression quantiles is proposed for constructing a fixed size confidence region of parameters of a linear model. The stopping time is asymptotically efficient and the confidence region is asymptotically consistent.