Statistical Method for combining multiple biomarkers

 
   Lead Investigator:    Eunhee Kim
   Institution      :    Brown University
   E-Mail           :    ekim@stat.brown.edu
   Proposal ID      :    280



   Proposal Description:
   One standard tool for assessing the accuracy of diagnostic tests is the Receiver Operating Characteristic (ROC) curve. Many methods have been developed to evaluate a single continuous-scale biomarker in the framework of ROC analysis. However, few statistical methods exist to accommodate multiple biomarkers in ROC analysis. In this project, we propose a method to combine continuous-scale biomarkers in order to optimize classification accuracy.  We will illustrate our method using NACC data.  Specifically, we will derive the optimal score for the multiple neuropsychological tests in 2005 and evaluate its accuracy to predict dementia.