Activity Number:
|
428
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 9, 2006 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistics in Epidemiology
|
Abstract - #305665 |
Title:
|
ROC Curve Analysis in Osteoporosis Screening
|
Author(s):
|
James Powers*+ and Margaret Gourlay and Kristine Ensrud
|
Companies:
|
The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill and VA Medical Center
|
Address:
|
Dept of Biostatistics, UNC-CH, Chapel Hill, NC, 27599,
|
Keywords:
|
ROC curve ; generalized linear model ; osteoporosis ; longitudinal study
|
Abstract:
|
Regression models for receiver operating characteristic (ROC) curve analysis, based on the theory of generalized linear models, have been proposed (Pepe 2000, Alonzo and Pepe 2002, Pepe and Cai 2004). These parametric distribution-free methods offer the ability to compare distributions of test results while accounting for important covariates. Correlation structures for clustered data also may be handled with this method. This poster applies ROC regression methods to bone density data from a large longitudinal observational study and discusses potential relevance to osteoporosis screening. Proposals for future statistical research are offered to supplement the data analysis.
|