SeminarsNonparametric Estimation For Time-Dependent AUC
271
reads
The area under the receiver operating characteristic curve (AUC) is the commonly used measure to evaluate or compare the predictive ability of markers to the disease status . Motivated by an angiographic coronary artery disease (CAD) study, our objective is mainly to compare the performance of different plasma markers in the prediction of CAD-related vital status at different time points within the study period. Instead of integrating the area under the estimated receiver operating characteristic curve (ROC) at a specified time point, a class of nonparametric estimators is proposed for the time-dependent AUC under the different censoring mechanisms. Moreover, the asymptotic properties of the estimators are developed and used to construct the approximated confidence regions of the AUCs and test for the performance of different markers. The finite sample properties and the usefulness of the estimators and procedures are investigated through extensive simulations. Finally, our proposed methods are applied to the illustrated empirical example.
reads
Chin-Tsang Chiang
2008-01-08
13:30:00 - 15:00:00
405 , Mathematics Research Center Building (ori. New Math. Bldg.)
The area under the receiver operating characteristic curve (AUC) is the commonly used measure to evaluate or compare the predictive ability of markers to the disease status . Motivated by an angiographic coronary artery disease (CAD) study, our objective is mainly to compare the performance of different plasma markers in the prediction of CAD-related vital status at different time points within the study period. Instead of integrating the area under the estimated receiver operating characteristic curve (ROC) at a specified time point, a class of nonparametric estimators is proposed for the time-dependent AUC under the different censoring mechanisms. Moreover, the asymptotic properties of the estimators are developed and used to construct the approximated confidence regions of the AUCs and test for the performance of different markers. The finite sample properties and the usefulness of the estimators and procedures are investigated through extensive simulations. Finally, our proposed methods are applied to the illustrated empirical example.