EViews.com

Referring to the eviews tutorial on the probit model, I'd like to ask a question on the final segment mentioned there concerning heteroskedasticity. First off, the example used a probit to model grade improvement and independent variables included gpa, tuce, and psi to refresh anyone's memory in case they haven't looked at it.

So the question is, why is it after transforming the independent variables (by multiplying them by fac) that psi and only psi is transformed again? This can be easily seen at the auxiliary regression step in the procedure, where it instructs us to run the regression of: brmr_y brmr_x (psi*(-xb)*fac). -- where brmr_x is a group already containing the transformed psi series.

hope this can be elaborated on by somebody, or at least is interesting to somebody

Thanks for reading

That is because, psi is also a regressor to the heteroscedastic variance equation. In the artificial regression, the terms following brmr_x are related to the variance part. Please see the cited reference (Davidson and MacKinnon, 1993) for more details.