squared error matrix comparisons between biased estimators An overview of recent results

Choosing the performance criterion to be mean squared error matrix, we have compared the least squares and Stein-rule estimators for coefficients in a linear regression model when the disturbances are not necessarily normally distributed. It is shown that none of the two estimators dominates the other, except in the trivial case of merely one regression coefficient where least squares is found to be superior in comparisons to Stein-rule estimators