Fully robust one-sided cross-validation for regression functions Academic Article uri icon

abstract

  • 2017, Springer-Verlag Berlin Heidelberg. Fully robust OSCV is a modification of the OSCV method that produces consistent bandwidths in the cases of smooth and nonsmooth regression functions. We propose the practical implementation of the method based on the robust cross-validation kernel HI in the case when the Gaussian kernel is used in computing the resulting regression estimate. The kernel HI produces practically unbiased bandwidths in the smooth and nonsmooth cases and performs adequately in the data examples. Negative tails of HI occasionally result in unacceptably wiggly OSCV curves in the neighborhood of zero. This problem can be resolved by selecting the bandwidth from the largest local minimum of the curve. Further search for the robust kernels with desired properties brought us to consider the quartic kernel for the cross-validation purposes. The quartic kernel is almost robust in the sense that in the nonsmooth case it substantially reduces the asymptotic relative bandwidth bias compared to . However, the quartic kernel is found to produce more variable bandwidths compared to . Nevertheless, the quartic kernel has an advantage of producing smoother OSCV curves compared to HI. A simplified scale-free version of the OSCV method based on a rescaled one-sided kernel is proposed.

published proceedings

  • COMPUTATIONAL STATISTICS

author list (cited authors)

  • Savchuk, O. Y., & Hart, J. D.

citation count

  • 3

complete list of authors

  • Savchuk, Olga Y||Hart, Jeffrey D

publication date

  • September 2017