A mean squared error criterion is used to compare five estimators of the coefficients in a linear regression model: least squares, principal components, ridge ...