Estimate the ROC curve and its area in the presence of nonignorable  
verification bias
 
   Lead Investigator:    Andrew Zhou
   Institution      :    NACC
   E-Mail           :    azhou@u.washington.edu
   Proposal ID      :    65


   Publications:


1  Chi YY, Zhou XH
   Receiver operating characteristic surfaces in the presence of verification bias
   Journal Article  J Appl Stat,57:1-23,2008,

   PUBLISHED



2  Harel O, Koepsell T, Kukull WA, Zhou XH
   Correcting for verification bias in the assessment of the accuracy of the clinical diagnosis of AD
   Abstract  Neurology,64:A69,2005,

   PUBLISHED



3  Harel O, Zhou XH
   Multiple imputation for the comparison of two screening tests in two-phase Alzheimer studies
   Journal Article  Stat Med,26:2370-2388,2007,

   PUBLISHED



4  Harel O, Zhou XH
   Multiple imputation: review of theory, implementation and software
   Journal Article  Stat Med,26:3057-3077,2007,

   PUBLISHED



5  Liu D, Zhou A
   A Model for Adjusting for Nonignorable Verification Bias in Estimation of ROC Curve and Its Area with Likelihood-Based Approach
   Journal Article  Biometrics,66:1119-1128,2011,PMC3618959

   PUBLISHED



6  Zhou XH , Duan X
   Quantile Regression-based Covariate-Specific ROC Curves
   2012   Abstract  WFN Annual Regional North American Meeting , 

   ACCEPTED





   Proposal Description:
   We proposed a likelihood-based approach to estimate the ROC curve and its area in the presence of nonignorable verification bias. We need a real example to demonstrate our approach. Here are some descriptions about the data set we need.  Sample size  The sample size needs to be greater than 3,000, or more preferably, greater than 5,000. Our simulation study suggests that a reasonably large sample size is required to estimate the nonignorable parameter.  Test result  The diagnostic test result needs to be continuous. Single test or multiple tests would both be fine.  Gold standard  The disease verification is partially available on the patients. Preferably, the verification proportion is about 50 to 90.