Experimental results on the auc of the bi-generalized exponential roc model using spread sheet functions

In medical diagnosis, testing of the performance of a particular diagnostic test is a common and good practice, after classifying the subjects in to different groups by various classification techniques, in particularly binary classification. Assessment of the performance of a diagnostic test can be achieved by Area under the Receiver Operating Characteristic Curve, simply denoted by AUC. For Diseased (D) and Healthy (H) normal populations, Bi-exponential model gives a closed form expression to the Area under the Curve (AUC). In this paper we report the results of simulated experiments on the properties of the AUC of the bi-generalized exponential ROC model. We first study the sensitivity of the AUC to changes in Scale parameter (λ) and Location parameter (µ). We will show that changes in λ do not alter the AUC of the model but with fixed scale parameter in D and H groups the AUC changes quickly when the distance between the location parameters is changed. Numerical illustrations for the proposed method are given with simulated data.