Semiparametric and Parametric Transformation Models for Comparing Diagnostic Markers with Paired Design

• Activity Number: 279
• Type: Contributed
• Date/Time: Wednesday, August 14, 2002 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology*
• Author(s): Kelly Zou*+ and W. Hall
• Affiliation(s): Harvard Medical School and University of Rochester Medical Center
• Keywords: sensitivity ; receiver operating characteristic ; Box-Cox transformation ; specificity ; area under the curve ; semiparametric method

Abstract

We develop two transformation models, namely semiparametric and parametric, for estimation and comparison of receiver-operating characteristic curves based on measurements derived from two diagnostic tests on the same subjects. We assume the existence of transformed measurement scales, one for each diagnostic test, on which the paired measurements have bivariate normal distributions. The resulting pair of ROC curves are estimated by maximum likelihood algorithms, using joint rank data in the semiparametric model with unspecified transformations, and using Box-Cox transformations in the parametric transformation case. Several hypothesis tests for comparing the two ROC curves, or characteristics of them, are developed. Two clinical diagnostic examples, on pancreatic cancer and on malignant melanoma, are presented. Simulation results are also provided.