Discriminant analysis is interchangeable with the formerly explained logistic regression approach. however, there are situations where the logistic regression is preferred over the discriminant analysis and vice versa. on the one hand, logistic regression is more robust in situations where the assumption that the set of independent variables is distributed multivariate normal with a common variance-covariance matrix is not satisfied. furthermore, logistic regression can handle both continuous as well as categorical independent variables, whereas discriminant analysis can only handle continuous variables. on the other hand, discriminant analysis can handle dependent variables with more than two categories, whereas a traditional binary logistic regression model can only handle a dependent variable with two categories. in this situation, the alternative to logistic regression when more than two categories are considered is called multinomial logistic regression.