The area Az under the ROC curve versus the number of features, k, used in linear discriminant analysis for Case 1 (identity covariance matrix). In this figure, it is assumed that an infinite number of features are available for classifier training, and that features i = 1,2,…,k are used for classification.