6:37 p.m.
On Fri, 8 Jul 2011, Francisco López Herrera wrote:
Maybe a SVAR gui design could be inspired on JMulti...Unfortunately
I
have not skills in C programming...With my best
regards...flh(Parameswara das)
On Fri, 8 Jul 2011, Henrique Andrade wrote:
Dear Jack, could you please take a look at the .jpg file attached?
It's
a simple proposal for the GUI implementation (more specifically a
suggestion for the matrix of restrictions).
Thank you for your suggestion, but I'm afraid it's not as simple as that:
JMulTi is (was?) a very clever UI written in java on top of GAUSS code. As
a consequence, Markus Krätzig was able to rely on the full power of Swing
for constructing the constraint matrices, whereas gretl package writers
can only rely on a limited subset of the GTK toolbox. (And no, I DON'T
think that should change --- at least in the near future.)
As for Henrique's proposal, it's a good starting point; true, we don't
have, at present, a user-level widget for modifying a matrix, but I guess
Allin could see to it (I couldn't: I simply can't do any GTK-related
coding, I'm hopeless at it).
But then, the problem is: do we really want the user to input the
constraint matrices (in implicit or explicit form)? These could be HUGE;
in an n-variate C-model, a constraint matrix in implicit form has n^2+1
columns and at least n*(n-1)/2 rows. If you had a system with 7 variables
(not unusual), would you really want to fill up a matrix with 50 columns
and 21 rows by hand? Or you could have the option of filling up the
restricted elements (the 0s and 1s), but then, what do you put in place of
an UNrestricted element? We may agree on some conventional value, like for
instance -999, but I doubt that would be considered particularly
intuitive. Moreover, in the case of a C-model we'd need to separate
somehow long-run from short-run restrictions which, again, I wouldn't know
how to do in a simple and clean way.
Perhaps the best candidate is a text box à la Eviews, in which you write
the restrictions in a hansl-like syntax. That would complicate the code
quite a bit, but I like the challenge. The only missing bit would be tyhe
possibility of having a resizable text box in our GUI interface to user
functions. Allin?
PS: I'm cc-ing the developer list, since this is getting a bit technical
IMO.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
10:20 a.m.
New subject: [Gretl-users] SVAR estimation
Am 08.07.2011 20:37, schrieb Riccardo (Jack) Lucchetti:
As for Henrique's proposal, it's a good starting point; true, we don't
have, at present, a user-level widget for modifying a matrix, but I
guess Allin could see to it (I couldn't: I simply can't do any
GTK-related coding, I'm hopeless at it).
I'm not sure that's true; you can define a matrix in the GUI already. It
should not be too difficult to invoke that matrix definition dialog
window from the function package call window, no? (To provide the input
of an matrix argument.)
But then, the problem is: do we really want the user to input the
constraint matrices (in implicit or explicit form)? These could be HUGE;
in an n-variate C-model, a constraint matrix in implicit form has n^2+1
columns and at least n*(n-1)/2 rows. If you had a system with 7
variables (not unusual), would you really want to fill up a matrix with
50 columns and 21 rows by hand? Or you could have the option of filling
up the restricted elements (the 0s and 1s), but then, what do you put in
place of an UNrestricted element? We may agree on some conventional
value, like for instance -999, but I doubt that would be considered
particularly intuitive. Moreover, in the case of a C-model we'd need to
separate somehow long-run from short-run restrictions which, again, I
wouldn't know how to do in a simple and clean way.
I agree that for some (most?) real-world applications it will not be
practical to do it via the GUI. However, for getting to know the
functionality I think a function package call GUI would be a big plus.
Also for teaching of course. And perhaps even for quite many smaller
applications.
So I think it should be possible in principle, and a user will see for
herself that at some point the GUI becomes impractical.
I'm ready to think this through and make suggestions. For example,
absent the possibility to put NAs in matrices (right?), I think your
-999 suggestion goes in the right direction. Gretl could prefill the
relevant matrices with that code (indicating unrestricted elements) and
then the user only has to change the restricted elements.
Perhaps the best candidate is a text box à la Eviews, in which you write
the restrictions in a hansl-like syntax. That would complicate the code
quite a bit, but I like the challenge. The only missing bit would be
tyhe possibility of having a resizable text box in our GUI interface to
user functions. Allin?
As I said, I think it's better to give the user various ways of
specifying the matrix arguments in specialized "widgets" (or the user
could simply pre-define the named matrix and at call time provide the
name, which is already possible now I think).
PS: I'm cc-ing the developer list, since this is getting a bit technical
IMO.
Yes.
thanks,
sven
2:50 a.m.
New subject: [Gretl-users] SVAR estimation
On Sat, 9 Jul 2011, Sven Schreiber wrote:
Am 08.07.2011 20:37, schrieb Riccardo (Jack) Lucchetti:
>
> As for Henrique's proposal, it's a good starting point; true, we don't
> have, at present, a user-level widget for modifying a matrix, but I
> guess Allin could see to it (I couldn't: I simply can't do any
> GTK-related coding, I'm hopeless at it).
I'm not sure that's true; you can define a matrix in the GUI already. It
should not be too difficult to invoke that matrix definition dialog
window from the function package call window, no? (To provide the input
of an matrix argument.)
Yes. If a packaged function has a matrix argument, a button
automatically appears beside the argument in the function-call
dialog which, when clicked, gives a GUI means of defining a
matrix.
However, I take Jack's point about the inherent difficulty of
building an SVAR specification via the GUI. Personally I'd consider
that pretty much insane, and I'd have no problem with introducing
the SVAR package as CLI-only.
Allin
8:51 a.m.
New subject: [Gretl-users] SVAR estimation
Am 11.07.2011 04:50, schrieb Allin Cottrell:
However, I take Jack's point about the inherent difficulty of
building an SVAR specification via the GUI. Personally I'd consider
that pretty much insane, and I'd have no problem with introducing
the SVAR package as CLI-only.
I'm still not convinced. Say for example you have a three-variable SVAR
and you want to do a B-model, and also only short-run restrictions. So
the user only has to specify the six zeroes in the B matrix. I really
think this should be possible via the GUI.
Having said that, maybe I should give it a try myself. So I would try to
package Jack's SVAR functions in a way that at least simple cases can be
done via the GUI. Let's see how that goes.
cheers,
sven
4918
days inactive
4921
days old
3 comments
3 participants
participants (3)
-
Allin Cottrell -
Riccardo (Jack) Lucchetti -
Sven Schreiber