The area under the receiver-operating characteristic (ROC) curve, AUC, is frequently used as diagnostic capacity measure and also as a goodness of fit index in logistic regression models. Given a considered (bio) marker, it measures the difference in the location parameters between two independent groups, each of them containing the positive and the negative subjects. Therefore, it can be interpreted as the effect size of the group on the studied variable (biomarker). In the present manuscript, we study the AUC interpretation within the two-sample problem. Both the non-parametric and the parametric tests join with the case where the marker is a categorical variable are considered. The use of the AUC as effect size provides a measure which is comparable and interpretable in different context: parametric, non-parametric and categorical cases are considered. Finally, in order to illustrate the problem, a real-world dataset is explored. [ABSTRACT FROM AUTHOR]