Re: [Gretl-users] HC3 and pvalues
On Mon, 24 Nov 2014, Ignacio Diaz-Emparanza wrote:
This is a question whose response I don't find in the manuals:
when you
estimate with OLS with the "robust standar errors" tickmarck active (for
example with HC3), I see the robust estimations of the coef. standard errors,
but where do the pvalues come from? are they from the normal (asymptotic)
distribution or are
they from a student-t ? and the pvalue of the F-statistic is from a F or a
chi-square distribution?
This needs documenting. At present the default is that such statistics
are referred to the student-t and F distributions. However, there's a
fairly new "set" variable named "robust_z" and if you do
set robust_z on
they will be referred to the normal and chi-square distributions.
Allin Cottrell