Activity Number:
|
362
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 14, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Statistical Computing*
|
Abstract - #301082 |
Title:
|
Permutation Testing for Classification and Regression Trees
|
Author(s):
|
Patrick Carmack*+ and Stephan Sain
|
Affiliation(s):
|
Southern Methodist University and University of Colorado, Denver
|
Address:
|
3225 Daniel Avenue, Rm 144 Heroy Bldg, Dallas, Texas, 75275-0332, USA
|
Keywords:
|
permutation test ; tree pruning ; classification and regression tree ; splitting criteria ; stopping criteria
|
Abstract:
|
Traditional approaches to building trees for classification and regression have relied on stopping rules, cost-complexity measures, and cross-validation to choose important splits and the size of the tree. An alternative is presented that is based an a simple permutation test gives information about the significance of individual splits, as well as the size of the tree. Comparisons with some of the traditional approaches will be considered with simulation and example.
|