|
Activity Number:
|
69
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 6, 2006 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #307320 |
|
Title:
|
Finite Elements Methods for Density Estimation
|
|
Author(s):
|
George Terrell*+
|
|
Companies:
|
Virginia Polytechnic Institute and State University
|
|
Address:
|
Statistics Department
Virginia Tech, Blacksburg, VA, 24061,
|
|
Keywords:
|
polynomial splines ; b-splines ; penalized least-squares ; ash methods
|
|
Abstract:
|
Nonparametric density estimates may be efficiently and rapidly computed and the results compactly stored using finite elements methods. These consist of polynomial splines estimated by least squares, which may then be smoothed by a combination of roughness penalties and direct averaging of the sort used in ASH methods. The results closely approximate popular classical methods such as kernels, but are generally faster to obtain. Boundary adjustments and variable smoothing are easy to include. Multivariate extensions are straightforward, using tensor-product splines.
|
