Multicolinearity

A second way to prevent multicolinearity is stated on the same paper, and its called the rule of thumb. This rule proposes to recognize the need to live with poorly conditioned non experimental data and looks to develop rules of thumb . Acceptable departures from orthogonality may be distinguished from harmful degrees of multicolinearity. Nevertheless Farrar´s and Glauber´s work has been widely criticized by C. Robert who has also proposed a better remedy for the problem called muticolinearity by computing the (n-2) nd order principal minors of each Xo´ Xo and comparing them with the corresponding Xoo´ Xooj and so on until the multicolinearity pattern is clear. A useful perspective in the way to prevent multicolinearity is proposed by Blanchard and it refers to “do nothing or incorporate additional information”. However this second statement sets the researcher into the probem of where to get data and what kind of adittional information should be added to prevent multicolinearity. Since we arised from the problem that data are never as much as the econometrician imagines.