7:22 p.m.
So far as I'm concerned we're about ready for the gretl 2019b
release. Does anyone have any bugs they want to report -- or do any
developers have new stuff they'd like to add -- before we do that?
If not I'll make a release in the next day or two.
Allin
2:14 a.m.
Am 08.05.2019 um 01:22 schrieb Allin Cottrell:
So far as I'm concerned we're about ready for the gretl 2019b
release.
Does anyone have any bugs they want to report -- or do any developers
have new stuff they'd like to add -- before we do that?
Spontaneously I remember the VIF stuff. There is a ticket on the tracker
about it: https://sourceforge.net/p/gretl/feature-requests/101/
with a suggestion from me two months ago.
I also contacted Lee about the situation in connection with his package
(and non-linear estimators I think), but have received no reply.
Then there's the thread "Lee's bkw.gfn and vif (was Re: (now
vif-related) Re: new accessor)", also in March.
It ended with you (Allin) saying "OK, we can try some experimentation."
thanks
sven
noon
Did I miss something? I take it that the bkw package has been "integrated"
into gretl making it unnecessary? That is fine with me. Is there something
I need to do?
Lee
On Wed, May 8, 2019 at 1:14 AM Sven Schreiber <svetosch(a)gmx.net> wrote:
Am 08.05.2019 um 01:22 schrieb Allin Cottrell:
> So far as I'm concerned we're about ready for the gretl 2019b release.
> Does anyone have any bugs they want to report -- or do any developers
> have new stuff they'd like to add -- before we do that?
Spontaneously I remember the VIF stuff. There is a ticket on the tracker
about it: https://sourceforge.net/p/gretl/feature-requests/101/
with a suggestion from me two months ago.
I also contacted Lee about the situation in connection with his package
(and non-linear estimators I think), but have received no reply.
Then there's the thread "Lee's bkw.gfn and vif (was Re: (now
vif-related) Re: new accessor)", also in March.
It ended with you (Allin) saying "OK, we can try some experimentation."
thanks
sven
_______________________________________________
Gretl-devel mailing list -- gretl-devel(a)gretlml.univpm.it
To unsubscribe send an email to gretl-devel-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-devel.gretlml.univpm.it/
--
Lee Adkins
Professor of Economics
Department Head, Economics
Oklahoma State University
Email: lee.adkins(a)okstate.edu
Fax: 405-744-5180
Phone: 405-744-5196
URL: www.learneconometrics.com
12:13 p.m.
On Wed, 8 May 2019, Lee Adkins wrote:
Did I miss something? I take it that the bkw package has been
"integrated"
into gretl making it unnecessary? That is fine with me. Is there something
I need to do?
I think you're right, Lee. The "vif" command now supports the new
$result accessor. As per the doc:
"Following this command the $result accessor may be used to retrieve a
bundle containing two matrices: a column vector holding the VIFs under
the key vif and the BKW matrix under BKW."
So I'm not sure there's anything left to do here.
Allin
> On Wed, May 8, 2019 at 1:14 AM Sven Schreiber <svetosch(a)gmx.net> wrote:
>
>> Am 08.05.2019 um 01:22 schrieb Allin Cottrell:
>>> So far as I'm concerned we're about ready for the gretl 2019b
release.
>>> Does anyone have any bugs they want to report -- or do any developers
>>> have new stuff they'd like to add -- before we do that?
>>
>> Spontaneously I remember the VIF stuff. There is a ticket on the tracker
>> about it: https://sourceforge.net/p/gretl/feature-requests/101/
>> with a suggestion from me two months ago.
>>
>> I also contacted Lee about the situation in connection with his package
>> (and non-linear estimators I think), but have received no reply.
>>
>> Then there's the thread "Lee's bkw.gfn and vif (was Re: (now
>> vif-related) Re: new accessor)", also in March.
>>
>> It ended with you (Allin) saying "OK, we can try some experimentation."
1:10 p.m.
Am 08.05.2019 um 18:13 schrieb Allin Cottrell:
On Wed, 8 May 2019, Lee Adkins wrote:
> Did I miss something? I take it that the bkw package has been
> "integrated"
> into gretl making it unnecessary? That is fine with me. Is there
> something
> I need to do?
This integration has indeed partly happened. However, ...
I think you're right, Lee. The "vif" command now
supports the new
$result accessor. As per the doc:
...last time I checked there were a bunch of estimation commands after
which 'vif' refused to work. In contrast, AFAIK Lee's package always
works whenever there's a $vcv matrix available. That was my reason as a
package manager not to retire Lee's package yet.
In March I had sent to this list a (possibly non-exhaustive) list of
non-working commands with 'vif':
* Doesn't work with vif:
- nls
- mle
- garch (presumably also arch)
- midasreg
- gmm
- biprobit
- tsls
- duration (a crash was fixed back then, not sure if it works now)
I haven't re-checked right now, has anything in this situation changed
since March?
thanks
sven
2:35 a.m.
On Wed, 8 May 2019, Sven Schreiber wrote:
In March I had sent to this list a (possibly non-exhaustive) list of
non-working commands with 'vif':
* Doesn't work with vif:
- nls
- mle
- garch (presumably also arch)
- midasreg
- gmm
- biprobit
- tsls
- duration (a crash was fixed back then, not sure if it works now)
IMO the new facility would be especially useful for the commands where the
user writes down her estimator herself (nls, gmm, mle). There can be cases
when you end setting up something that is very weakly identified (or
downright unidentified) without you noticing, and the result is BFGS going
on and on and on, spewing out something that is utter nonsense, if
anything.
In those cases, I've found the spectral decomposition of $vcv to be a
very handy tool. Trivial example:
<hansl>
set seed 9999
nulldata 100
x = normal()
u = 0.001*uniform()
a = 0
b = 0
mle llik = -0.5*e^2
e = x - (a+b*u)
params a b
end mle
matrix eigenvec
lam = eigensym($vcv, &eigenvec)
print lam eigenvec
</hansl>
In this case, the model is just OLS, but b is very weakly identified
because the regressor's variance is ridiculous. The problem is easy to
spot if you look at the second eigenvector.
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
11:33 a.m.
Am 09.05.2019 um 08:35 schrieb Riccardo (Jack) Lucchetti:
On Wed, 8 May 2019, Sven Schreiber wrote:
> In March I had sent to this list a (possibly non-exhaustive) list of
> non-working commands with 'vif':
>
> * Doesn't work with vif:
> - nls
> - mle
> ...
IMO the new facility would be especially useful for the commands where
the user writes down her estimator herself (nls, gmm, mle).
What I'd like to know in the present context, does Lee's bkw package
help you there? That would be an argument to keep it around a bit more.
Or are you pointing out features that are missing everywhere so far?
thanks
sven
1:12 p.m.
On Thu, 9 May 2019, Sven Schreiber wrote:
> IMO the new facility would be especially useful for the commands
> where
> the user writes down her estimator herself (nls, gmm, mle).
What I'd like to know in the present context, does Lee's bkw package
help you there? That would be an argument to keep it around a bit
more.
Yes, that's what I mean.
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
2:57 p.m.
On Thu, 9 May 2019, Riccardo (Jack) Lucchetti wrote:
On Thu, 9 May 2019, Sven Schreiber wrote:
> > IMO the new facility would be especially useful for the commands where
> > the user writes down her estimator herself (nls, gmm, mle).
>
> What I'd like to know in the present context, does Lee's bkw package
> help you there? That would be an argument to keep it around a bit more.
Yes, that's what I mean.
Lee's bkw package has the advantage, as things stand, of giving you
the BKW matrix with just a covariance matrix and a list of parameter
names required as arguments.
The "vif" command as it stands works off the last estimated model, and
is somewhat restrictive in the sorts of model that it handles. That's
primarily because of the VIF part as such, which wants a plain old
list of regressors, as opposed to the BKW part which just needs a
covariance matrix.
I agree with Sven that some clean-up and enhancement is needed, but I
suggest we propose that till after the 2019b release. I think what
might be best, when we get to it, is
* Revert the "vif" command to just producing the VIFs (which are not
very sophisticated but are easy to interpret). And support a larger
set of estimators to the extent possible (but this would not include
nls, mle or gmm). The VIF vector to be available as $result.
* Add a built-in bkw function which uses what's now in the C code to
replace Lee's package. Any covariance matrix accepted, so nls, mle and
gmm should be fine.
* Rejig the GUI "Collinearity" menu item so that it produces what's
appropriate for the model in the window -- possibly with a choice of
VIFs, BKW or both, if applicable.
* I think the doc should have something substantive (besides the
reference to Lee's paper, which is fine in itself) by way of
explanation of the BKW matrix, which is IMO by no means transparent.
Allin
4:21 p.m.
Am 09.05.2019 um 20:57 schrieb Allin Cottrell:
I agree with Sven that some clean-up and enhancement is needed, but
I
suggest we propose that till after the 2019b release. I think what
might be best, when we get to it, is
* Revert the "vif" command to just producing the VIFs (which are not
very sophisticated but are easy to interpret). And support a larger
set of estimators to the extent possible (but this would not include
nls, mle or gmm). The VIF vector to be available as $result.
* Add a built-in bkw function which uses what's now in the C code to
replace Lee's package. Any covariance matrix accepted, so nls, mle and
gmm should be fine.
* Rejig the GUI "Collinearity" menu item so that it produces what's
appropriate for the model in the window -- possibly with a choice of
VIFs, BKW or both, if applicable.
* I think the doc should have something substantive (besides the
reference to Lee's paper, which is fine in itself) by way of
explanation of the BKW matrix, which is IMO by no means transparent.
Sounds good to me! (Also the postponement.)
thanks
sven
4:37 p.m.
On Thu, 9 May 2019, Sven Schreiber wrote:
Am 09.05.2019 um 20:57 schrieb Allin Cottrell:
> I agree with Sven that some clean-up and enhancement is needed, but I
> suggest we propose that till after the 2019b release. I think what
> might be best, when we get to it, is
>
> * Revert the "vif" command to just producing the VIFs (which are not
> very sophisticated but are easy to interpret). And support a larger
> set of estimators to the extent possible (but this would not include
> nls, mle or gmm). The VIF vector to be available as $result.
>
> * Add a built-in bkw function which uses what's now in the C code to
> replace Lee's package. Any covariance matrix accepted, so nls, mle and
> gmm should be fine.
>
> * Rejig the GUI "Collinearity" menu item so that it produces what's
> appropriate for the model in the window -- possibly with a choice of
> VIFs, BKW or both, if applicable.
>
> * I think the doc should have something substantive (besides the
> reference to Lee's paper, which is fine in itself) by way of
> explanation of the BKW matrix, which is IMO by no means transparent.
Sounds good to me! (Also the postponement.)
I like this too.
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
5:30 p.m.
On Thu, 9 May 2019, Riccardo (Jack) Lucchetti wrote:
On Thu, 9 May 2019, Sven Schreiber wrote:
> Am 09.05.2019 um 20:57 schrieb Allin Cottrell:
>
>> I agree with Sven that some clean-up and enhancement is needed, but I
>> suggest we propose that till after the 2019b release. I think what
>> might be best, when we get to it, is
>>
>> * Revert the "vif" command to just producing the VIFs (which are not
>> very sophisticated but are easy to interpret). And support a larger
>> set of estimators to the extent possible (but this would not include
>> nls, mle or gmm). The VIF vector to be available as $result.
>>
>> * Add a built-in bkw function which uses what's now in the C code to
>> replace Lee's package. Any covariance matrix accepted, so nls, mle and
>> gmm should be fine.
>>
>> * Rejig the GUI "Collinearity" menu item so that it produces
what's
>> appropriate for the model in the window -- possibly with a choice of
>> VIFs, BKW or both, if applicable.
>>
>> * I think the doc should have something substantive (besides the
>> reference to Lee's paper, which is fine in itself) by way of
>> explanation of the BKW matrix, which is IMO by no means transparent.
>
> Sounds good to me! (Also the postponement.)
I like this too.
Great, seems like we're on the same page. Adding a built-in bkw
function should be quick work, the other points may take a little
longer.
Allin
1906
days inactive
1908
days old
11 comments
4 participants
participants (4)
-
Allin Cottrell -
Lee Adkins -
Riccardo (Jack) Lucchetti -
Sven Schreiber