11:10 p.m.
Most of the time, I fail to reconstitute the sessions I have saved. I
sometimes manage to open them, but I don't understand what is different from
when I can't. Am I missing something about it ?
Thanks
1:55 a.m.
On Mon, 23 Jan 2006, Arnaud Bervas wrote:
Most of the time, I fail to reconstitute the sessions I have
saved. I sometimes manage to open them, but I don't understand
what is different from when I can't. Am I missing something about
it ?
Sorry, I have to admit this aspect of gretl is potentially flaky.
In my experience things work OK if you simply (a) carry out some
tasks in gretl; (b) save your session; and (c) re-open it. If
things are not working for you in ths context, please post a
description of what's happening.
Problems start arising when you make changes to a saved session and
re-save it -- the main problem here being the logical design issue
of what *ought* to happen in this situation. It's harder than you
might think. Any thoughts on this would be appreciated --
preferably, informed by an examination of what gretl writes into a
session (*.gretl) file.
Allin Cottrell
2:06 a.m.
New subject: sessions reconstitution [2]
On Mon, 23 Jan 2006, Arnaud Bervas wrote:
Most of the time, I fail to reconstitute the sessions I have saved.
A second thought: Does anyone on the list happen to know how R's
mechanism for saving the binary image of one's workspace works? It
may be that a radical redesign of gretl's "session saving" mechanism
is in order, along the lines of doing some sort of binary memory
dump.
Allin Cottrell
7:10 a.m.
New subject: sessions reconstitution [2]
Thanx for that Allin.
Now I have another problem with exogenous variables. In VAR module (and in
VECM too) there is no possibility to specify number of lags of exogenous
variables. They are allways included with zero lags. I also tried with
creating lags manually and then to include them as exogenous but it's not
working. They do not apper in VAR and VECM modules (This is possible in
OLS). This means that Gretl threats exogenous as deterministics i.e. as
constant and trend i.e. without lags.
Is it possible to do that somehow?
If not, then I am opting for this option as a new feature! This is realy an
usefull thing.
For Jack: I tried to do AIC and BIC tests for lag selection in VAR but the
procedure stops here
matrix mincrit = { 1.0E20 , 1.0E20 }
I think the problem is in numbers in brackets. As I am from Croatia, I tried
to change Regional settings on Italian and English but the procedure still
remains unfinished. I'll try again, maybe I missed something. Do you have an
idea?
It will be great to include it as a feature in next version.
From: Allin Cottrell <cottrell(a)wfu.edu>
Reply-To: Gretl list <gretl-users(a)ricardo.ecn.wfu.edu>
To: Gretl list <gretl-users(a)ricardo.ecn.wfu.edu>
Subject: Re: [Gretl-users] sessions reconstitution [2]
Date: Mon, 23 Jan 2006 21:06:59 -0500 (EST)
On Mon, 23 Jan 2006, Arnaud Bervas wrote:
>Most of the time, I fail to reconstitute the sessions I have saved.
A second thought: Does anyone on the list happen to know how R's mechanism
for saving the binary image of one's workspace works? It may be that a
radical redesign of gretl's "session saving" mechanism is in order, along
the lines of doing some sort of binary memory dump.
Allin Cottrell
_______________________________________________
Gretl-users mailing list
Gretl-users(a)ricardo.ecn.wfu.edu
http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users
_________________________________________________________________
Are you using the latest version of MSN Messenger? Download MSN Messenger
7.5 today! http://messenger.msn.co.uk
7:36 a.m.
New subject: sessions reconstitution [2]
On Tue, 24 Jan 2006, john w wrote:
Thanx for that Allin.
Now I have another problem with exogenous variables. In VAR module (and in
VECM too) there is no possibility to specify number of lags of exogenous
variables. They are allways included with zero lags. I also tried with
creating lags manually and then to include them as exogenous but it's not
working. They do not apper in VAR and VECM modules (This is possible in
OLS). This means that Gretl threats exogenous as deterministics i.e. as
constant and trend i.e. without lags.
Is it possible to do that somehow?
If not, then I am opting for this option as a new feature! This is realy an
usefull thing.
Short-run workaround: define lagged variables using another name. For
example
genr xlag1 = x(-1)
or if you feel adventurous
loop for i=1..4
genr xlag$i = x(-$i)
end loop
In the long run, I still think this could be solved by setting up tools
for handling lists in the GUI interface, but that will have to wait a bit.
For Jack: I tried to do AIC and BIC tests for lag selection in VAR
but the
procedure stops here
matrix mincrit = { 1.0E20 , 1.0E20 }
I think the problem is in numbers in brackets. As I am from Croatia, I tried
to change Regional settings on Italian and English but the procedure still
remains unfinished. I'll try again, maybe I missed something. Do you have an
idea?
Sorry, I was assuming you're running the CVS version. If you're using
Windows, that should work with the nightly snapshot, which is basically
CVS with a 1-day lag. The matrix infrastructure is totally new and
the script cannot work with 1.5.0. OTOH, there is no real need for
matrices in that script, so modifying it to avoid matrix constructs
should be a relatively simple task.
It will be great to include it as a feature in next version.
We'll think about it.
Riccardo `Jack' Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
jack(a)dea.unian.it
http://www.econ.univpm.it/lucchetti
4:55 p.m.
New subject: VAR dialog (was Re: sessions reconstitution [2])
On Tue, 24 Jan 2006, jack quoted john w:
> Now I have another problem with exogenous variables. In VAR
module (and in
> VECM too) there is no possibility to specify number of lags of exogenous
> variables. They are allways included with zero lags. I also tried with
> creating lags manually and then to include them as exogenous but it's not
> working. They do not apper in VAR and VECM modules (This is possible in
> OLS). This means that Gretl threats exogenous as deterministics i.e. as
> constant and trend i.e. without lags.
The VAR command is one of the older interfaces in the program, and
it could do with some updating. In CVS I have added a "--trend"
option to the command, with an associated checkbox, "Include a
trend", in the VAR dialog box. I have also the renamed the upper
right-hand variable selection box to "Exogenous" rather than
"Deterministic" variables.
However, as you point out, including lags of exogenous variables is
problematic. In the original design, the upper box was indeed
intended for non-stochastic variables (trend, dummies), where the
inclusion of lags would not make sense. And lagged variables are
not shown in the left-hand list, for the obvious reason that this
could get very confusing, since lags of the endogenous variables are
handled automatically.
Jack suggests:
Short-run workaround: define lagged variables using another name.
For
example
genr xlag1 = x(-1)
Yes, this is a good idea (for the present). You can thereby fool
gretl into thinking that "xlag1" is not really a lag, and it will be
displayed in the left-hand variable selection list.
In the long run, I still think this could be solved by setting up
tools for handling lists in the GUI interface, but that will have
to wait a bit.
Agreed, on both counts.
Quting john w again:
> I tried to do AIC and BIC tests for lag selection in VAR but the
> procedure stops here
> matrix mincrit = { 1.0E20 , 1.0E20 }
Jack:
Sorry, I was assuming you're running the CVS version. If
you're
using Windows, that should work with the nightly snapshot, which
is basically CVS with a 1-day lag. The matrix infrastructure is
totally new and the script cannot work with 1.5.0.
The Windows snapshot usually follows CVS with a short lag, but
there's some discretion here and in fact I haven't made a snapshot
lately since there's a lot of new material in CVS that could be
unstable. There will be a new snapshot as soon as CVS has settled
down (shouldn't be long now).
Allin Cottrell
8:41 a.m.
New subject: VAR dialog (was Re: sessions reconstitution [2])
Thnx for accepting my advice. I hope that VAR and VECM module will be
finished soon.
Please could you send me a link to download the latest CVS version for
Windows.
I can't find it on the page anymore.
From: Allin Cottrell <cottrell(a)wfu.edu>
Reply-To: Gretl list <gretl-users(a)ricardo.ecn.wfu.edu>
To: jack(a)deanovell.unian.it, Gretl list <gretl-users(a)ricardo.ecn.wfu.edu>
Subject: VAR dialog (was Re: [Gretl-users] sessions reconstitution [2])
Date: Tue, 24 Jan 2006 11:55:17 -0500 (EST)
On Tue, 24 Jan 2006, jack quoted john w:
>>Now I have another problem with exogenous variables. In VAR module (and
>>in
>>VECM too) there is no possibility to specify number of lags of exogenous
>>variables. They are allways included with zero lags. I also tried with
>>creating lags manually and then to include them as exogenous but it's not
>>working. They do not apper in VAR and VECM modules (This is possible in
>>OLS). This means that Gretl threats exogenous as deterministics i.e. as
>>constant and trend i.e. without lags.
The VAR command is one of the older interfaces in the program, and it could
do with some updating. In CVS I have added a "--trend" option to the
command, with an associated checkbox, "Include a trend", in the VAR dialog
box. I have also the renamed the upper right-hand variable selection box
to "Exogenous" rather than "Deterministic" variables.
However, as you point out, including lags of exogenous variables is
problematic. In the original design, the upper box was indeed intended for
non-stochastic variables (trend, dummies), where the inclusion of lags
would not make sense. And lagged variables are not shown in the left-hand
list, for the obvious reason that this could get very confusing, since lags
of the endogenous variables are handled automatically.
Jack suggests:
>Short-run workaround: define lagged variables using another name. For
>example
>
> genr xlag1 = x(-1)
Yes, this is a good idea (for the present). You can thereby fool gretl
into thinking that "xlag1" is not really a lag, and it will be displayed in
the left-hand variable selection list.
>In the long run, I still think this could be solved by setting up tools
>for handling lists in the GUI interface, but that will have to wait a bit.
Agreed, on both counts.
Quting john w again:
>>I tried to do AIC and BIC tests for lag selection in VAR but the
>>procedure stops here
>>matrix mincrit = { 1.0E20 , 1.0E20 }
Jack:
>Sorry, I was assuming you're running the CVS version. If you're using
>Windows, that should work with the nightly snapshot, which is basically
>CVS with a 1-day lag. The matrix infrastructure is totally new and the
>script cannot work with 1.5.0.
The Windows snapshot usually follows CVS with a short lag, but there's some
discretion here and in fact I haven't made a snapshot lately since there's
a lot of new material in CVS that could be unstable. There will be a new
snapshot as soon as CVS has settled down (shouldn't be long now).
Allin Cottrell
_______________________________________________
Gretl-users mailing list
Gretl-users(a)ricardo.ecn.wfu.edu
http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users
_________________________________________________________________
Are you using the latest version of MSN Messenger? Download MSN Messenger
7.5 today! http://messenger.msn.co.uk
8:10 a.m.
New subject: Building gretl under windows: volunteers?
Something that's been floating around in my mind for a while: most people
who actively work on the gretl development are Linux users: Allin,
Ignacio, Cristian, myself, for instance, are all linux addicts. On the
other hand, I have the feeling that most gretl users use windows as their
primary environment (this is certainly true of my students).
In some cases, this is a nuisance: once a Windows-only user spots a bug,
he has no other choice but to report it, hoping it will be fixed soon,
even if he had all the necessary skills to fix it himself. That's because
up to now no-one (as far as I know) has ever tried to compile gretl under
windows.
Building gretl under windows is certainly possible. Unfortunately, I have
no experience whatsoever with the GNU toolchain for Win32 (nor have I
intention of setting up a Windows box for trying, no thanks); but I know
for a fact that all the programs you need to build gretl (gcc compiler,
make etc.) are available under windows; one implementation (but there may
be others) can be found at http://cygwin.org.
Has anyone on the list some experience with this?
Riccardo `Jack' Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
jack(a)dea.unian.it
http://www.econ.univpm.it/lucchetti
1:19 p.m.
New subject: Building gretl under windows: volunteers?
jack schrieb:
Something that's been floating around in my mind for a while:
most people
who actively work on the gretl development are Linux users: Allin,
Ignacio, Cristian, myself, for instance, are all linux addicts. On the
other hand, I have the feeling that most gretl users use windows as their
primary environment (this is certainly true of my students).
In some cases, this is a nuisance: once a Windows-only user spots a bug,
he has no other choice but to report it, hoping it will be fixed soon,
even if he had all the necessary skills to fix it himself. That's because
up to now no-one (as far as I know) has ever tried to compile gretl under
windows.
Building gretl under windows is certainly possible. Unfortunately, I have
no experience whatsoever with the GNU toolchain for Win32 (nor have I
intention of setting up a Windows box for trying, no thanks); but I know
for a fact that all the programs you need to build gretl (gcc compiler,
make etc.) are available under windows; one implementation (but there may
be others) can be found at http://cygwin.org.
Has anyone on the list some experience with this?
Very limited only, I have mingw (not cygwin) installed and built something successfully by
following
exact instructions. Due to the translation job I also have the cvs-repository here, so I
could try
if you provide those instructions.
But what exactly would the purpose be? You guys seem very successful at cross-compiling
for Windows,
no?
-sven
1:49 p.m.
New subject: Building gretl under windows: volunteers?
On Tue, 24 Jan 2006, Sven Schreiber wrote:
Very limited only, I have mingw (not cygwin) installed and built
something successfully by following
exact instructions. Due to the translation job I also have the cvs-repository here, so I
could try
if you provide those instructions.
But what exactly would the purpose be? You guys seem very successful at cross-compiling
for Windows,
no?
Sorry, I wasn't quite clear. The purpose is not to have a windows
executable; we've got that. The purpose is assembling a recipe for
compiling the source under windows.
Suppose you only use windows and there's something you'd like to change in
the source (recent example: "exogenous" vs "deterministic" in VARs).
That's a trivial change, so you think "Instead of bugging Allin, I'll do
it myself". You spot the line(s) to change, you change it, and then what?
If you can't compile, you have no way of checking if your changes have the
effects you wanted or, worse, break something. So eventually, your patch
sits idle.
Since the majority of our users use windows, I think that would be a
sizeable step forward. (It'd be better still if the majority used linux,
but I digress... and using win is a crime that contains its own punishment
anyway)
Riccardo `Jack' Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
jack(a)dea.unian.it
http://www.econ.univpm.it/lucchetti
2:01 p.m.
New subject: Building gretl under windows: volunteers?
jack schrieb:
Sorry, I wasn't quite clear. The purpose is not to have a windows
executable; we've got that. The purpose is assembling a recipe for
compiling the source under windows.
Ah I see. I will try it, but I'm expecting to not get it working right away, and I
don't have the
time for long trial-and-error sessions with makefiles and such, sorry.
However, there may be a workaround: Windows users can run Linux as a virtual machine with
the free
vmware player. There are pre-built virtual Linux machines available linked from the vm
website; I'm
currently in the process of trying such a thing out. At least some of those distros should
come with
complete compilers, which would make it possible for Windows users to build gretl on Linux
the same
way as you do, for free (and legally).
-sven
2:08 p.m.
New subject: Building gretl under windows: volunteers?
On Tue, 24 Jan 2006, Sven Schreiber wrote:
jack schrieb:
>
> Sorry, I wasn't quite clear. The purpose is not to have a windows
> executable; we've got that. The purpose is assembling a recipe for
> compiling the source under windows.
>
Ah I see. I will try it, but I'm expecting to not get it working right away, and I
don't have the
time for long trial-and-error sessions with makefiles and such, sorry.
However, there may be a workaround: Windows users can run Linux as a virtual machine with
the free
vmware player. There are pre-built virtual Linux machines available linked from the vm
website; I'm
currently in the process of trying such a thing out. At least some of those distros
should come with
complete compilers, which would make it possible for Windows users to build gretl on
Linux the same
way as you do, for free (and legally).
I don't know about the performace of vmware. In any case, installing a
whole distro on a virtual machine for the sole purpose of compiling gretl
seems a bit overkill. But, if you have to, choose Debian :-)
Just out of curiosity: assuming you have mingw installed and working, what
happens if you run the "configure" script in the top directory of the
source from a bash shell? Does it run at all? Does it complain about
missing stuff?
Riccardo `Jack' Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
jack(a)dea.unian.it
http://www.econ.univpm.it/lucchetti
5:05 p.m.
New subject: Building gretl under windows: volunteers?
On Tue, 24 Jan 2006, jack wrote:
Sorry, I wasn't quite clear. The purpose is not to have a
windows
executable; we've got that. The purpose is assembling a recipe for
compiling the source under windows...
I think this would be useful.
At present the gretl "configure" script is not going to work on
Windows, even with a fully-endowed mingw installation, because it
makes various *nix-specific assumptions.
It would not be too difficult to fix that by adding some appropriate
conditionals. I don't have time to work on that right now, but if
anyone's interested I think a reasonable approach would be to
cross-reference gretl's "configure.in" with the Makefile and
"config.mk" in the win32 subdirectory of the gretl source (which are
used for the cross-compile).
Allin Cottrell
7:25 a.m.
New subject: sessions reconstitution [2]
On Mon, 23 Jan 2006, Allin Cottrell wrote:
On Mon, 23 Jan 2006, Arnaud Bervas wrote:
> Most of the time, I fail to reconstitute the sessions I have saved.
A second thought: Does anyone on the list happen to know how R's
mechanism for saving the binary image of one's workspace works? It
may be that a radical redesign of gretl's "session saving" mechanism
is in order, along the lines of doing some sort of binary memory
dump.
Once the "bundles" mechanism Allin and I have been discussing is properly
in place, this should be much easier. All we need to settle is a precise
list of what has to be carried over between sessions.
Riccardo `Jack' Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
jack(a)dea.unian.it
http://www.econ.univpm.it/lucchetti
6667
days inactive
6669
days old
13 comments
5 participants
participants (5)
-
Allin Cottrell -
Arnaud Bervas -
jack -
john w -
Sven Schreiber