|
Activity Number:
|
419
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #300096 |
|
Title:
|
Evaluation of Prognostic Models Using Area Between Curves (ABC)
|
|
Author(s):
|
Mithat Gonen*+
|
|
Companies:
|
Memorial Sloan-Kettering Cancer Center
|
|
Address:
|
307 East 63rd street 3rd Floor, New York, NY, 10065,
|
|
Keywords:
|
predictive accuracy ; discrimination ; calibration ; censored data ; local regression
|
|
Abstract:
|
Evaluation of prognostic models usually involve measuring the discriminatory power using the ROC curve and the AUC as well as an assessment of calibration by comparing predictions with observations. The methods developed in the context of binary data and logistic regression require some adjustments when used in presence of censored data. This talk will discuss the use of ROC curves with censored data and computation of the AUC with Cox models as well as with Kaplan-Meier estimates. Assessment of calibration will be performed by an estimate of the calibration curve using local regression. Area between curves (ABC) will also be discussed as a measure of calibration of prognostic models. Staging of colon cancer using national databases will be used as an example.
|
