|
Activity Number:
|
92
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #308744 |
|
Title:
|
Evaluating Predictive Capacity of Continuous Biomarkers
|
|
Author(s):
|
Ying Huang*+ and Margaret Pepe and Ziding Feng
|
|
Companies:
|
University of Washington and Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center
|
|
Address:
|
Department of Biostatistics, Seattle, WA, 98195,
|
|
Keywords:
|
Predictiveness Curve ; risk prediction ; ROC curve ; cross-sectional studies
|
|
Abstract:
|
Statistical methodology is needed to critically evaluate biomarkers. A well-established criterion for marker evaluation is classification accuracy, often characterized by the ROC curve. However, classification is not always the goal. Oftentimes we use markers to predict risk of disease. Since the criteria for evaluating risk prediction markers are different from those for classification markers, we suggest an alternative to the ROC curve for their evaluation. We propose using the predictiveness curve to display the distribution of risk predicted by markers. This tool is valuable for health policy makers who are interested in policy analysis of screening and referral applied to the whole population. We propose an estimator for making inference about the curve and for making pointwise comparisons between curves in cross-sectional studies. We also develop estimators for subpopulations.
|
