|
Activity Number:
|
311
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 14, 2002 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology*
|
| Abstract - #301809 |
|
Title:
|
Modified Recursive Partitioning Methods for Cost-effective Classifications
|
|
Author(s):
|
Hua Jin*+ and Ying Lu+
|
|
Affiliation(s):
|
University of California, San Francisco and University of California, San Francisco
|
|
Address:
|
530 Parnassus Avenue , San Francisco, California, 94143-1290, USA 530 Parnassus Avenue , San Francisco, California, 94143-1290, USA
|
|
Keywords:
|
Recursive partitioning ; optimum split ; equivalent split ; sensitivity and specificity ; COST-EFFECTIVE CLASSIFICATIONS ; osteoporosis research
|
|
Abstract:
|
Recursive partitioning (RP) methods have been used to develop classification rules for separating subjects into subgroups that have different risk profiles for a disease. Conventional RP consists of splitting and pruning steps. The cost differences of risk factors are not considered in the splitting step. In this paper, we present an algorithm with some modifications. First, we group risk factors into different cost categories, such as clinical, laboratory, and radiology, etc. At each splitting step, the optimum split based on the conventional RP algorithm is compared to splits based on other risk factors, including cost categories and variables used in previous splits. An equivalent split is selected based on an equivalence test for both sensitivity and specificity (Lu, et al., 2002). The tree continues to grow until its full development. There is no modification of the pruning algorithm. The final equivalent tree is compared to the optimum tree using the same tests but with more stringent thresholds. An example of its application in osteoporosis research will be presented.
|
- 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 2002 program |
