Comparison of Discriminant Analysis and Logistic Regression Analysis: An Application on Caesarean Births and Natural Births Data

The discriminant analysis DA and the logistic regression analysis LRA are two statistical techniques used for analyzing data and predicting group membership from a set of predictors. Many applications have been done in this area. In this paper, we have been focus for the comparison of the two statistical techniques through applying on real data and exactly on caesarean births and natural births data. The comparison was depending on two statistical criteria; apparent error rate AER and apparent correct classification rate ACCR and performing stepwise procedure.

The results of the analysis showed that the performance of both techniques gave high ability in discriminating the kind of birth, whereas DA slightly exceeds LRA in the apparent correct classification rate and performed better than LRA in the births data. On the other hand, the results of DA showed that out of ten predicted variables, seven predictors exhibited strong evidence in classifying and discriminating the kind of birth, while the results of LRA showed that six predicted variables out of ten predictors have contributed significantly to discriminate the kind of birth. The suitable model for both techniques has been estimated depending on the selected predictors.