SFB 303 Discussion Paper No. A - 235

Author: Härdle, Wolfgang, and Raymond J. Carroll
Title: Biased Crossvalidation for a Kernel Regression Estimator and its Derivatives
Abstract: For univariate nonparametric regression, we compute the mean squared error of a kernel regression estimator and its derivatives (Gasser and M¨ller, 1984), extending slightly the conditions of applicability of this estimator. We show how to estimate this mean squared error and thus the best smoothing parameter by what Scott and Terrell (1987) call biased crossvalidation, which is essentially a refined version of the "plug-in" method. This bandwidth estimator is shown to be asymptotically optimal in the sense of Härdle and Marron (1985).
Keywords: Nonparametric regression, kernel regression, bandwidth selection, bias correction, mean squared error
JEL-Classification-Number: 211
Creation-Date: April 1989
Unfortunately this paper is not available online. Please contact us to order a hardcopy.

SFB 303 Homepage

19.05.1998, Webmaster