discriminate analysis and logistic regression existence of multicolleniarty problem(Empirical Study on Anemia)

Abstract

Abstract The method binery logistic regression and linear discrimint function of the most important statistical methods used in the classification and prediction when the data of the kind of binery (0,1) you can not use the normal regression therefore resort to binary logistic regression and linear discriminant function in the case of two group in the case of a Multicollinearity problem between the data (the data containing high correlation) It became not possible to use binary logistic regression and linear discriminant function, to solve this problem, we resort to Partial least square regression. In this, search the comparison between binary logistic regression and linear discriminant function using error Category. In the practical side in the collection of data on the data on anemia collection Two variables are severe anemia (0) and and chronic anemia (1) and several variables about the disease. The Data were collected from several Iraqi hospitals, where samples collected from patients at the hospital are asleep, and previous cases lay in the hospital a sample of (140) the patient is infected with the disease. When the test data and found that Multicollinearity problem, It has been processed using a method partial least square. The research found that linear discriminant function It is the best in the classification of data from binary logistic regression classified as linear discriminant function the data correctly and more accurate than binary logistic regression.