ABSTRACT
This chapter starts with the general introduction of the receiver operating characteristic (ROC) curves and then it discusses the construction of the empirical likelihood (EL) statistic for the nonparametric estimator of the whole or partial area under the ROC curve (AUC) that has a form of the U-statistic. It discusses the best combinations of multiple biomarkers utilizing ROC curve analysis. An important task of constructing the EL statistic is to incorporate the correct variance estimate to the EL statistic as discussed in Chapter 7. The problem is that the typical variance formula for U-statistics is inaccurate to estimate the variability of the estimator as plug-in estimators of the quantiles are used. In this context, this chapter provides details of a correct variance estimation strategy. Relevant R codes are provided for practitioners.
