5 p.m.
I am attaching a test script. For case 3 it works, but not for case 4, and as Sven said,
nor for case 2.
Kind regards,
Andreas.
P.S. In what concerns editing the plots, I did not find anything to change the color or he
area in the GUI plot editor.
5:56 p.m.
Am 02.07.2018 um 19:00 schrieb Andreas Zervas:
I am attaching a test script. For case 3 it works, but not for case
4,
and as Sven said, nor for case 2.
Ok, thanks. If there are no further complications, I think I should be
able to deal with this tomorrow.
P.S. In what concerns editing the plots, I did not find anything to
change the color or he area in the GUI plot editor.
I'll have a look, but it's not super high priority :-)
cheers,
sven
12:35 a.m.
On Mon, 2 Jul 2018, Sven Schreiber wrote:
Am 02.07.2018 um 19:00 schrieb Andreas Zervas:
> I am attaching a test script. For case 3 it works, but not for case 4, and
> as Sven said, nor for case 2.
>
Ok, thanks. If there are no further complications, I think I should be able
to deal with this tomorrow.
Quick diagnosis: In cases 2 and 4 the function SVAR_coint is exiting
early on error, because jbeta has 5 rows and the function thinks it
ought to have 6. So on exit the "SVARobj" has not been filled with
the required matrices. Then when we get to the function vecm_est,
jbeta is 0 x 0 and the attempt to extract a non-null submatrix
necessarily fails ("Index value 1 is out of bounds").
I've taken the liberty of making SVAR_coint fail with an error
message (in git) if jbeta is deemed not to be of the correct
dimensions. Of course feel free to revert that, but it seems some
more error-reporting is needed.
Allin
8:56 a.m.
Am 03.07.2018 um 02:35 schrieb Allin Cottrell:
Quick diagnosis: In cases 2 and 4 the function SVAR_coint is exiting
early on error, because jbeta has 5 rows and the function thinks it
ought to have 6. So on exit the "SVARobj" has not been filled with the
required matrices. Then when we get to the function vecm_est, jbeta is
0 x 0 and the attempt to extract a non-null submatrix necessarily
fails ("Index value 1 is out of bounds").
Hi Allin, I don't think this is what the (original) error is about, or
I'm not following you. Could it be that you used the test script but
didn't change the --uc option in the vecm line when you chose cases 2 or 4?
I've taken the liberty of making SVAR_coint fail with an error message
(in git) if jbeta is deemed not to be of the correct dimensions. Of
course feel free to revert that, but it seems some more
error-reporting is needed.
More error reporting is always good!
thanks,
sven
12:10 p.m.
On Tue, 3 Jul 2018, Sven Schreiber wrote:
Am 03.07.2018 um 02:35 schrieb Allin Cottrell:
>
> Quick diagnosis: In cases 2 and 4 the function SVAR_coint is exiting early
> on error, because jbeta has 5 rows and the function thinks it ought to have
> 6. So on exit the "SVARobj" has not been filled with the required matrices.
> Then when we get to the function vecm_est, jbeta is 0 x 0 and the attempt
> to extract a non-null submatrix necessarily fails ("Index value 1 is out of
> bounds").
Hi Allin, I don't think this is what the (original) error is about, or I'm
not following you. Could it be that you used the test script but didn't
change the --uc option in the vecm line when you chose cases 2 or 4?
Yes, that's exactly what I did. So it seems I'm not talking about the
original error. Nonetheless, this too seems like an error condition
that merits attention.
> I've taken the liberty of making SVAR_coint fail with an
error message (in
> git) if jbeta is deemed not to be of the correct dimensions. Of course feel
> free to revert that, but it seems some more error-reporting is needed.
More error reporting is always good!
Before I made that change in git, SVAR_coint just silently returned
non-zero on error. On a quick inspection of the SVAR doc, I don't
think it's clear that the user has to check for non-zero return values
from SVAR functions (on pain of getting incomprehensible errors from
subsequent function calls). In fact, the alphabetical listing of SVAR
functions doesn't mention their return types.
I'd suggest that you and Jack consider using "funcerr" wherever
relevant. Of course the design choice is yours, but if the choice is
just to flag errors with a non-zero return, that should be made clear
in the doc and the example scripts should illustrate error-checking,
as in
err = SVAR_coint(<args>)
if err
...
else
...
endif
Allin
11:14 a.m.
Am 02.07.2018 um 19:56 schrieb Sven Schreiber:
Am 02.07.2018 um 19:00 schrieb Andreas Zervas:
> I am attaching a test script. For case 3 it works, but not for case
> 4, and as Sven said, nor for case 2.
>
Ok, thanks. If there are no further complications, I think I should be
able to deal with this tomorrow.
It seems my hunch was correct, it was simply a bug in the matrix
multiplication order, so nobody before used an SVEC with a rank higher
than one and with the restricted deterministics setup -- or at least
nobody before cared to report the problem...
In vecm_est() the following two lines needed to be changed:
matrix mu = jbeta[n+1,] * jalpha' # was: jalpha * jbeta[n+1,]
...
matrix mu = olspar[1,] | (jbeta[n+1,] * jalpha') # was: (jalpha *
jbeta[n+1,])'
It works for me, but I'm attaching a new gfn for testing -- but beware
this is not the official updated release, for example the sample files
and the help is missing there. Also attached is a slightly generalized
test script.
I'm not currently at a machine where the git access is set up, so
pushing this to git would have to wait until tonight (unless Jack does
it first).
cheers,
sven
11:19 a.m.
Am 03.07.2018 um 13:14 schrieb Sven Schreiber:
I'm not currently at a machine where the git access is set up,
BTW Allin, the current SVAR version is 1.35 and this is apparently also
shipped with the snapshots, but when I go (on a June 21 snapshot) to
Help / Look for addons, the result is "SVAR 1.34, up to date".
cheers,
sven
12:14 p.m.
On Tue, 3 Jul 2018, Sven Schreiber wrote:
Am 03.07.2018 um 13:14 schrieb Sven Schreiber:
>
> I'm not currently at a machine where the git access is set up,
BTW Allin, the current SVAR version is 1.35 and this is apparently also
shipped with the snapshots, but when I go (on a June 21 snapshot) to Help /
Look for addons, the result is "SVAR 1.34, up to date".
Oops, that's now corrected.
Allin
12:30 p.m.
Am 03.07.2018 um 14:14 schrieb Allin Cottrell:
On Tue, 3 Jul 2018, Sven Schreiber wrote:
> BTW Allin, the current SVAR version is 1.35 and this is apparently
> also shipped with the snapshots, but when I go (on a June 21
> snapshot) to Help / Look for addons, the result is "SVAR 1.34, up to
> date".
Oops, that's now corrected.
In the meantime we're at 1.36 due to the corrected bug, could it be
updated again? (Right now I'm seeing 1.35 as up to date.)
thanks,
sven
11:12 a.m.
Am 04.07.2018 um 14:30 schrieb Sven Schreiber:
Am 03.07.2018 um 14:14 schrieb Allin Cottrell:
> On Tue, 3 Jul 2018, Sven Schreiber wrote:
>
>> BTW Allin, the current SVAR version is 1.35 and this is apparently
>> also shipped with the snapshots, but when I go (on a June 21
>> snapshot) to Help / Look for addons, the result is "SVAR 1.34, up to
>> date".
>
> Oops, that's now corrected.
>
In the meantime we're at 1.36 due to the corrected bug, could it be
updated again? (Right now I'm seeing 1.35 as up to date.)
Andreas (OP), I think version 1.36 is now directly availabile. (Help /
Look for Addons) Could you update to 1.36 and test whether it also works
for you?
Thanks,
sven
3:22 p.m.
On Tue, 3 Jul 2018, Sven Schreiber wrote:
I'm not currently at a machine where the git access is set up, so
pushing
this to git would have to wait until tonight (unless Jack does it first).
Done; I also bumped up the version number.
-------------------------------------------------------
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
-------------------------------------------------------
3:28 p.m.
Am 03.07.2018 um 17:22 schrieb Riccardo (Jack) Lucchetti:
On Tue, 3 Jul 2018, Sven Schreiber wrote:
> I'm not currently at a machine where the git access is set up, so
> pushing this to git would have to wait until tonight (unless Jack
> does it first).
Done; I also bumped up the version number.
Thanks!
-sven
2362
days inactive
2366
days old
11 comments
4 participants
participants (4)
-
Allin Cottrell -
Andreas Zervas -
Riccardo (Jack) Lucchetti -
Sven Schreiber