Journal of Statistics: Advances in Theory and ApplicationsAuthor's: JAN KOLÃÄŒEK and IVANKA HOROVÃ
Pages: [91] - [103]
Received Date: November 1, 2012
Submitted by:
The aim of the contribution is to extend the idea of an iterative method known for a kernel density estimate to kernel regression. The method is based on a suitable estimate of the mean integrated square error. This approach leads to an iterative quadratically convergent process. We conduct a simulation study comparing the proposed method with the well-known cross-validation method. Results are implemented in Matlab.
kernel regression, bandwidth selection, iterative method.
