On Dealing with Multicollinearity Using New Hybridized Regression Model

Abstract

Background Multicollinearity is a serious problem to be considered while performing the multiple regression analysis and so also the multiple logistic regressions. The presence of Multicollinearity leads to several consequences which makes the purpose of regression analysis futile. To address the problem of Multicollinearity, several remedies like Ridge regression or Factor Analysis are suggested. Methods This article describes an alternative procedure which is potential enough to deal with Multicollinearity which encompasses two different statistical procedures, viz., Statistical Discriminant Analysis and Logistic Regression, and hence may be called Hybridized-LR. Results The suggested method shows high potentiality in terms of accuracy measures when compared with all other algorithms, which is substantiated by the application based on the data collected on chronic kidney disease.