Mathematical Programming Approach for Variable Selection in Discriminant Analysis: An Environmental Application

Abstract

The environmental impact assessment of projects is based on several variables. It is desirable to select the most important variables needed for classifying any project in only one of three classes according to severity of possible environmental impact. In this study, the selection of variables in discriminant analysis between more than two groups using the Mathematical programming approach is applied to Egypt environmental impact assessment survey 2000 data to select the most important variables for classifying new projects in the true classes. The results are compared with those of the stepwise method. The comparison shows that according to the percent of correct classification, the Mathematical programming model is better than the stepwise method