A Comparison of Parameters Estimation Methods for the Negative Binomial Regression Model under Multicollinearity Problem by Using Simulation

Abstract

This study discussed a biased estimator of the Negative Binomial Regression model known as (Liu Estimator), This estimate was used to reduce variance and overcome the problem Multicollinearity between explanatory variables, Some estimates were used such as Ridge Regression and Maximum Likelihood Estimators, This research aims at the theoretical comparisons between the new estimator (Liu Estimator) and the estimators of Maximum Likelihood (ML) and Ridge Regression (RR) by using the mean square error (MSE) criterion, where the variance of the Maximum Likelihood (ML) comes in the presence of the problem Multicollinearity between the explanatory variables. In this study, the Monte Carlo simulation was designed to evaluate the performance of estimations using the criterion for comparison, the mean square error (MSE). The simulation results showed important an estimated Liu and superior to the RR and MLE estimator Where the number of explanatory variables is (p=5) and the sample size is (n=100), where the number of explanatory variables is (p=3) and for all sizes, and also when (p=5) for all sizes except size (n=100), the RR regression method is the best.