ABSTRACT
In Chapters 3-5 we have studied series, kernel and smoothing spline estimators which represent what are arguably the most popular approaches to nonparametric regression. In this final chapter we study another way to use spline functions for data smoothing: namely, least-squares splines. These estimators have local fitting qualities similar to those for kernel and smoothing spline estimators. However, least-squares splines do not admit kernel or series representations, even asymptotically, which distinguishes them from the smoothing methods of previous chapters.
