To complete the design, it is necessary to define how εi is generated. We consider
two cases. As it is standard in the analysis of the performance of robust estimators,
we conduct some experiments in which εi is obtained from a contaminated normal.
how εi is generated