© 1990 by Biometrika Trust
MISCELLANEA |
Why bandwidth selectors tend to choose smaller bandwidths, and a remedy
Department of Statistics, Colorado State University Fort Collins, Colorado 80523, U.S.A.
Received for publication 1 February 1989.
Revision received 1 May 1989.
| Abstract |
|---|
This note is concerned with the problem of automatic data-driven bandwidth selectors for nonparametric regression. Though some selectors were shown to be consistent and asymptotically unbiased by Rice (1984) and Hãrdle, Hall & Marron (1988), it is often observed in simulation studies that most selectors are biased toward undersmoothing. An explanation for this is given. The source of the large sample variation in the bandwidth estimates is also pointed out. This leads to the consideration of a new procedure which is a simple modification of a classical selector. A simulation study suggests that the proposed selector is much more consistent than the classical one.
Key Words: Bandwidth selection Fourier transform Kernel estimate Nonparametric regression Periodogram