Title of Publication: “ON THE ESTIMATION OF EXPECTED ERROR RATES IN DISCRIMINANT ANALYSIS”
Author(s): G.A. OSUJI, E.O. OBODO, S.I. ONYEAGU
Year of Publication: 2011
When a sample discriminant function DS is computed, it is desired to estimate the chance of misclassification using DS. This is often done by classifying the sample with the help of DS or by computing Ø(-1/2 D), where Ø is the cumulative normal distribution, and D2 is Mahalanobi’s distance. When D2 is applied to new simulated sample(s), we observed that the probabilities of misclassification are usually found to be greater than those computed from the initial samples.