|
Activity Number:
|
227
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #305152 |
|
Title:
|
Double-Semiparametric ROC Regression Analysis
|
|
Author(s):
|
Xiao-Hua Andrew Zhou*+ and Huazhen Lin
|
|
Companies:
|
University of Washington and University of Washington
|
|
Address:
|
Box 357232, Seattle, WA, 98195-7232,
|
|
Keywords:
|
non-parametric ; ROC curves ; diagnostic accuracy ; transformation models
|
|
Abstract:
|
ROC regression methodology offers an opportunity to investigate how factors affect test accuracy. For directly modeling ROC curves, both parametric methods---where the link and baseline functions are specified---and semiparametric methods---where the link function is specified but the baseline function is unspecified---have been developed. However, the misspecification of either the link or the baseline function can lead to substantial bias for the ROC curve estimates. In this talk, we extend the existing direct ROC regression models to allow arbitrary nonparametric link and baseline functions. We show the proposed estimators for the regression parameters and ROC curves are asymptotically normal and consistent with the parametric convergent rate, n^{-1/2}. We illustrate our approach with a real dataset.
|
