Using Some Robust Methods For Handling the Problem of Multicollinearity

Abstract

The multiple linear regression model is an important regression model that has attracted many researchers in different fields including applied mathematics, business, medicine, and social sciences , Linear regression models involving a large number of independent variables are poorly performing due to large variation and lead to inaccurate conclusions , One of the most important problems in the regression analysis is the multicollinearity Problem, which is considered one of the most important problems that has become known to many researchers  , As well as their effects on the multiple linear regression model, In addition to multicollinearity, the problem of outliers in data is one of the difficulties in constructing the regression model , Leading to adverse changes when taking linear regression as a basis for hypothesis testing .

In this paper, we present some robust methods for estimating the parameters of the multiple linear regression model, a ridge regression method for based on the LTS estimator and Liu method for based on the LTS estimator, Using the simulation, these two methods were compared according to the mean squares error (MSE) , The comparison showed that the Liu-LTS method is the best in estimating the parameters of the multiple linear regression model.