A Comparison of Logistic Regression and Linear Discriminant Analysis in Predicting of Female Students Attrition from School in Bangladesh

Abstract

This study aimed to compare the predictive accuracy and also the classification accuracy of two models using real data of school attrition. The overall classification accuracy for both models was determined by the classification accuracy rate. Logistic regression analysis (LRA) and linear discriminant analysis (LDA) classified 78.33% and 78.38% of girls respectively, in-school and out-of-school correctly. The AUROC curve for LRA was 80.63%, while it was 80.57% for the LDA. The LRA has sensitivity and specificity were 45.81% and 91.60%, respectively, and the LDA had a sensitivity of 46.81% and specificity of 91.01%. The overall classification rate for both was good. In comparison with the conventional LRA model, the LDA was better than LRA in the correct classification rate. In general, the LRA model looks appropriate for prediction accuracy while LDA seems suitable to be used for classification techniques.