Dear professor Schreiber,
As may you already saw, my message with the picture made the mail list far later because as you suspected, the image was more than 100kb.
My main interest is the long run and the loading coefficients because I am investigating mainly long run causality. However, most papers applying PMG report the mean group short run coefficients. In fact, I don't remember any in my field not reporting the mean group short run coefficients. In addition, I may get some relevant outcomes which can improve a little bit the quality of the papers as well.
But, I really understand your point because I always questioned the pertinence of those mean group coefficients as almost always we get a "salad" of positive and negative significant coefficients and also positive and negative insignificant ones.
Of course, I can do it in eviews, but I would be very happy if I can migrate from Eviews. Indeed, as I mentioned some weeks ago, the PMG package has the very desirable Hausman test and the MG alternative for the long run coefficients.
Kind regards,
Reynaldo
On 3/24/21, 10:40 AM Sven Schreiber <svetosch@gmx.net> wrote:
Am 24.03.2021 um 15:19 schrieb Riccardo (Jack) Lucchetti:
> On Wed, 24 Mar 2021, Reynaldo Senra wrote:
>> Very much thank you for your quick responses. Attached is a
>> screenshot with Eviews and Gretl.
Your message didn't make it to the list, it seems. Perhaps the
screenshot was too large; anyway, I haven't seen it.
>> As you will see, everything is identical with the long run and
>> loading coefficients. However, the"fullcoeffs" output only
>> has identical coefficients for the long run estimates but I dont know
>> which are short run estimates because no coefficients are similar to
>> the ones in Eviews and "fullcoeffs" does not specify which are short
>> and long
>> run coefficients (of course I know which are long run because of the
>> output). Indeed, apparently, "fullcoeffs" does not report the
>> standard errors.
The documentation in section 1.1 explains: "The ordering is: loading –
(homogeneous) long-run coeffs – lagged
differences – exogenous (including constant)." However, I admit that for
mtype 1 and 2 also the contemporaneous differences of the x-es enter the
picture, and the ordering of the y and x differences inside is not
self-explanatory. So perhaps we could improve the doc in that respect.
Also, you have to make sure the setup is really identical, for example
as regards mtype.
It is true that we don't provide the standard errors, although of course
it wouldn't be difficult. But again, what do you need them for? - An
honest question.
>>
>> Maybe the way the sort run coefficients are not obtained by OLS in
>> Eviews. I don't know, everything is quite confusing for me.
>
>
> I have the feeling (but I may well be wrong) that Eviews uses the PMG
> method for estimating the long-run parameters and the loadings and
> then runs a pooled model for producing the short-run parameters.
No no, the short run is heterogeneous of course. I'm quite sure Eview
also just runs separate second-stage OLS regressions, because that's the
best thing one can do.
We need to see the exact specifications I guess. I'm pretty sure I
cross-checked with Eviews for the package update.
cheers
sven
_______________________________________________
Gretl-users mailing list -- gretl-users@gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave@gretlml.univpm.it
Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/