|
Activity Number:
|
416
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #305612 |
|
Title:
|
Nonparametric Bayesian Bootstrap in ROC Curve Study
|
|
Author(s):
|
Jiezhun Gu*+ and Subhashis Ghosal
|
|
Companies:
|
North Carolina State University and North Carolina State University
|
|
Address:
|
2501 Founders Drive, Raleigh, NC, 27695,
|
|
Keywords:
|
Bayesian bootstrap ; asymptotic properties ; ROC curve
|
|
Abstract:
|
Receiver operating characteristic (ROC) curve is applied widely in measuring discriminatory ability of diagnostic or prognostic tests, leading to parametric and nonparametric estimation methods. In this paper, we present the Bayesian bootstrap method to estimate ROC curves. Integrated absolute error of ROC curves is introduced as a measure of accuracy especially useful in simulation studies to evaluate the performance of different methods of estimation of the ROC curves and the estimate of the area under the curve. Flexibility and computational simplicity are the two main advantages of the Bayesian bootstrap. We also study the asymptotic properties of the Bayesian bootstrap estimate, providing a justification of this method.
|
