Abstract

The accuracy of discrimination between the two populations namely healthy and diseased in a medical diagnosis can be assessed through the renowned statistical technique called Receiver Operating Characteristic (ROC) curve. The Area Under the ROC Curve (AUC) is the traditional index to measure the diagnostic accuracy. Several parametric distributions are assumed to plot a parametric ROC curve viz. Normal, Exponential, Gamma, Lognormal, Rayleigh, etc. But in all these cases, single distribution is assumed for both the populations. This paper deals with the problem of estimating ROC , AUC and standard error of AUC based on healthy test scores follow Normal distribution and diseased test scores follow Exponential distribution, we call it as Normal-Exponential ROC curve. The proposed model is explored using simulation as well as real life example.

Author: Sudesh Pundir and R. Amala

Received on: September, 2013

Accepted on: March, 2014