|
Activity Number:
|
362
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #305671 |
|
Title:
|
Supplementing a Validation Test Sample
|
|
Author(s):
|
Frank W. Samuelson*+
|
|
Companies:
|
FDA
|
|
Address:
|
10903 New Hampshire Ave, Silver Spring, MD, 20993-0002,
|
|
Keywords:
|
learning machines ; test ; validation ; supplement
|
|
Abstract:
|
Requiring an independent, external test sample is standard for obtaining unbiased performance validation of diagnostic learning machines. Learning machines evolve over time as developers improve their performance and repeated validation of these learning machines may be necessary. Repeated use of a fixed validation test sample will lead to bias and optimistic estimates, but obtaining an entirely new test sample with every validation may not be possible. This presentation suggests a method for supplementing a test sample with additional cases such that the performance of the learning machine on repeated use of that test sample must have limited bias, given certain assumptions. This limit is used to construct a criterion for the size of a supplement for a test sample that is modest and depends only upon the size of the original test sample.
|
- 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 2009 program |
