|
Activity Number:
|
47
|
|
Type:
|
Invited
|
|
Date/Time:
|
Sunday, August 6, 2006 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #304962 |
|
Title:
|
Ensemble Models for Risk Prediction with Survey and Multilevel Data
|
|
Author(s):
|
Stuart A. Gansky*+ and Nancy F. Cheng
|
|
Companies:
|
University of California, San Francisco and University of California, San Francisco
|
|
Address:
|
3333 California Street, Suite 495, San Francisco, CA, 94118,
|
|
Keywords:
|
classification and regression tree ; random forests ; survey ; nesting ; logit ; hybrid
|
|
Abstract:
|
Classification and regression tree (CART), multiple additive regression tree (MART), and random forests (RF) prediction models are powerful prediction techniques. Multilevel or complex survey sample data need to have their designs taken into account to provide proper inference. Ensemble models to account for these designs with a multilevel hybrid CART survey logit model approach and differential misclassification costs were developed. Extensions for ensemble models with MART and with RF are proposed. A stratified cluster sample health examination survey of early childhood caries is used for illustration. Results show that this hybrid approach may be effective at identifying associations with health disparities. Support: US DHHS NIH/NIDCR, NCMHD U54DE14251.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2006 program |
