Generalized maximum entropy estimation of linear regression model with multicollinearity problem

In this study, the linear regression model y = + X u b is considered. The explanatory variables matrix X may subject to multicollinearity in applications. In the presence of collinearity, classical estimators produce incorrect results. Biased but stable estimators can be used to overcome this problem. In this study, generalized maximum entropy (GME) estimator is compared with the ordinary least squares (OLS) and ridge estimators according to the mean squared error (MSE) criteria. For this purpose, California poverty data set (Ramanathan, 2002) is analysed. In the application section, the bootstrap method is used to obtain MSE values. In conclusion, GME estimator is decided as the best estimator. Keywords: General linear model, multicollinearity, generalized maximum entropy, bootstrap