Superiority of the Stochastic Restricted Liu Estimator under misspecification

Abstract

This paper deals with the use of correct prior infromation in the estimation of regres-sion coefficients when the regression model is misspecified due to the exclusion of some relevant regressor variables. In particular, the attention is focused on the Stochastic Re-stricted Liu estimator introduced by Hubert and Wijekoon (2004), which outperforms Liu estimator with respect to the matrix mean squared error matrix criterion. Further the su-periority of the Stochastic Restricted Liu predictor over the Liu predictor is also exam-ined, and concluded that there are situations where the Stochastic Restricted Liu predictor outperforms the Liu predictor with respect to the mean squared error matrix criterion even the model is misspecified.