Simulation study on the area under the ROC curve in the presence of missingness
  
   Lead Investigator:    Hunyong Cho
   Institution      :    Loyola University Chicago
   E-Mail           :    hunyong.cho@gmail.com
   Proposal ID      :    732


   Publications:


1  Cho H, Harel O, Mattews GJ
   Confidence intervals for the area under the receiver operating characteristic curve in the presence of ignorable missing data
   Journal Article  Int Stat Rev,87:152-177,2019,PMC6472951

   PUBLISHED





   Proposal Description:
   My coworkers and I are examining combinations of confidence interval methods for the the area under the ROC curve (AUC) and multiple imputation techniques through simulations. The confidence interval methods we are examining are all Wald-type intervals including those proposed in Hanley  McNeil (1982), DeLong et al (1988), Bamber (1975), and Newcombe (2006). The multiple imputation techniques considered are fully conditional specification (implemented using the MICE package in R) and NORM. We compare the performances of the intervals in our simulations using metrics such as coverage probability and interval length. While the simulations are useful in comparing these methods, we believe the strength of our manuscript will be greatly increased by presenting a real data example in our article.