COMPARISON OF MLE, LASSO, AND LIU ESTIMATOR METHODS TO OVERCOME MULTICOLLINEARITY IN MULTINOMIAL LOGISTIC REGRESSION: SIMULATION STUDY
The purpose of this study is to compare the performance of the MLE, LASSO, and Liu Estimator, and MLE methods in dealing with multicollinearity using simulated data with n=50,75,150, and 300 in multinomial logistic model (p=6) with and 0,99. The best model was compered using AIC, MSE, BIC.is the best based zn the MSE, SE, AIC, BIC values. The result showed that LASSO and Liu Estimator methods were able to overcome partial multicollinearity in 3 independent variables and full multicollinearity 6 independent variables much better than MLE method. This result is based on MSE, AIC, and BIC values of LASSO and Liu Estimator which are much smaller than those of MLE.
Keywords– LASSO, Liu Estimator, MLE, multicollinearity
