I think you'll find that the probit MLE itself is inconsistent for the
coefficients when the underlying errors are heteroskedastic and/or not
independent. Hence, the estimates themselves are not useful. Fixing
standard errors for inconsistent coefficient estimates is not really
something you'd want to do....
On Tue, Apr 23, 2013 at 12:39 PM, Reynald Majetti <
Is there a way to compute robust standard errors for potential
autocorrelation problems in probit models ?
Gretl-devel mailing list
Professor of Economics