Robustness or Efficiency, A Test to Solve the Dilemma

When dealing with the presence of outliers in a dataset, the problem of choosing between the classical ordinary least squares and robust regression methods is sometimes addressed inadequately. In this article, we propose using a Hausman-type test to determine whether a robust S- estimator is more appropriate than an ordinary least squares one in a multiple linear regression framework, on the basis of the trade-off betewen robustness and efficiency. An economic example is provided to illustrate the usefulness of the test.