A short note on optimal bandwidth selection for kernel estimators
For kernel estimators of an unknown density it has been proposed to use a data-based bandwidth which is as close to the (nonrandom, unknown) minimiser of the Mean Integrated Squared Error as possible. In this note it will be shown that this strategy of bandwidth selection is not optimal.
Saved in:
| Main Author: | |
|---|---|
| Format: | Article (Journal) |
| Language: | English |
| Published: |
1990
|
| In: |
Statistics & probability letters
Year: 1990, Volume: 9, Issue: 1, Pages: 23-25 |
| DOI: | 10.1016/0167-7152(90)90090-T |
| Online Access: | Verlag, Volltext: http://dx.doi.org/10.1016/0167-7152(90)90090-T Verlag, Volltext: http://www.sciencedirect.com/science/article/pii/016771529090090T 
|
| Author Notes: | Enno Mammen |
| Summary: | For kernel estimators of an unknown density it has been proposed to use a data-based bandwidth which is as close to the (nonrandom, unknown) minimiser of the Mean Integrated Squared Error as possible. In this note it will be shown that this strategy of bandwidth selection is not optimal. |
|---|---|
| Item Description: | Available online 22 March 2002 Gesehen am 01.03.2018 |
| Physical Description: | Online Resource |
| DOI: | 10.1016/0167-7152(90)90090-T |