Online Program

Within cluster resampling method on clustered ROC data

*Zhuang Miao, George Mason University 
Liansheng Larry Tang, George Mason University 

Keywords: ROC, Clustered data, resampling

Clustered ROC data is a type of data that each cluster has several diseased and nondiseased observations. Within the same cluster, observations are naturally correlated, and the cluster sizes may be informative of the patient’s disease status. The traditional ROC methods on clustered data could result large bias and lead to incorrect statistical inference. We introduce within-clustered resampling methods on clustered ROC data to account for the within cluster correlation and informative cluster sizes. The within cluster resampling ROC method works as follows. First, one observation is randomly selected from each cluster, then the traditional ROC methods are applied on the resampled data to obtain ROC estimates. These steps are performed multiple times and the average of resampled ROC estimates is the final estimator. The proposed method does not require a specific within cluster correlation structure and yields valid estimator when the cluster size are informative. We prove the asymptotic properties of the proposed method, and compare the proposed estimator to existing methods in extensive simulation studies.