Abstract
When using kernel-based estimators for probability density functions, a key problem is the choice of smoothing parameter. All existing procedures for this choice suffer from some significant drawback. We propose a simple location-dependent procedure for the choice of smoothing parameters. Initial comparisons with two standard choice methods, due to Parzen and to Breiman et al., with respect to the mean percent error (MPE), show that, while all three methods have comparable performance in estimating unimodal densities, ours and Breiman's both outperform Parzen's in multimodal cases. Since our method is considerably simpler than Breiman's, we develop it further. In particular, through an approximation of the L 1 criterion, we develop a distribution-free goodness-of-fit criterion for evaluating our choice of parameter.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.