Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
1148684 | Journal of Statistical Planning and Inference | 2012 | 8 Pages |
Abstract
In this paper we propose and study a new kernel regression estimator in which the kernel is taken from a properly adapted location-scale family of the design distribution. We show that, while the original smoothing may be performed with sub-optimal bandwidths, adaptation of proper scale parameters yields overall optimal estimators. Unlike traditional smoothing methodology, our approach does not aim at estimating pivotal higher order derivatives.
Related Topics
Physical Sciences and Engineering
Mathematics
Applied Mathematics
Authors
Gerrit Eichner, Winfried Stute,