[Gretl-users] Re: The 2021c release candidate is available (with apologies for cross-posting)
Am 19.08.2021 um 15:40 schrieb Alecos Papadopoulos:
In the User's guide ch. 22 I cannot find a mention of the
heteroskedasticity-robust estimator "HC3a", which one can find as an
option in the model menu.
Thanks for pointing out the gap in the manual. The built-in reference
(when you hit the help button in the HCCME tab in the preferences
dialog) has this to say (excerpt):
"HC0 produces the original “White's standard errors”; HC1, HC2, HC3 and
HC3a are subsequent variations that are generally reckoned to produce
superior (more reliable) results. For details of the estimators, see
MacKinnon and White (Journal of Econometrics, 1985) or Davidson and
MacKinnon, Econometric Theory and Methods (Oxford, 2004). The labels
given here are those used by Davidson and MacKinnon. Variant “HC3a” is
the jackknife, as described in MacKinnon and White; HC3 is a close
approximation to the jackknife. "
Could you perhaps include this also in the list of the
"HCx" esitmators
presented I think in p. 196?
Page numbers are fragile because there are US letter and A4 versions of
the same document, apart from translated versions perhaps. I think
you're talking about the end of section 22.2 which for me is on p. 189.
Yes, the list should be completed. (However, I don't know the exact
difference between HC3 and HC3a offhand.)
thanks
sven