While the specificity and sensitivity of a diagnostic test depend on the cutoff value chosen, a useful summary measure to consider is the area under the ROC curve. It can be shown mathematically that the area under curve is P(β0TXD>β0TXD¯) (Bamber [15]). Under a binormal probability model, Su and Liu [12] showed that this quantity is optimized using the linear discriminant function. This motivates our choice of consideration of these variables. We next present an algorithm for estimation of these functions.
