An Asymptotically Efficient Solution to the Bandwidth Problem of Kernel Density Estimation. Revision.
Abstract
A data-driven method of choosing the bandwidth, h, of a kernel density estimator is proposed. It is seen that this means of selecting h is asymptotically equivalent to taken the h that minimizes a certain weighted version of the mean integrated square error. Thus, for a given kernel function, the bandwidth can be chosen optimally without making precise smoothness assumptions on the underlying density. The proposed technique is a modification of cross-validation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1984
- Accession Number
- ADA149600
Entities
People
- J. S. Marron
Organizations
- University of North Carolina at Chapel Hill