9:14 p.m.
OK, there's now a snapshot with some of the new features we've been
discussing. It has matrix methods, 15-character variable names, and
a new lag-selection dialog box.
This version still needs to have the bugs shaken out of it, so I
wouldn't recommend feeding it to students yet. It is available at
this URL only:
http://ricardo.ecn.wfu.edu/pub/gretl/gretl_install.exe
--
Allin Cottrell
Department of Economics
Wake Forest University, NC
9:12 p.m.
New subject: New Windows snapshot - PERFECT!
All I can say Allin: it's PERFECT!
Yes, that's it!
Thank you.
I got a few advices:
1. VAR module output: It will be great to split endogenous variables and
exogenous variables with putting deterministics between them. To get clearer
output.
2. VAR and VECM module impulse responses graph: it will be better to plot
confidence intervals with lines in some other colour rather to use fans.
This is to get clearer output. It is a mess when plotting combined graph.
The plot window is small and can't be extended so when plotting CI as fans
you can't see any more the main impulse response graph. Other software have
lines too.
3. Pleeeaaassee put AIC and BIC tests for lag selection in VAR and VECM. For
example in Tests module of VAR and VECM.
4. In VAR roots plot output: it is good have numerical values of the roots
in this plot but when some root is on or outside the circle you can't see
root values because plot window can't be extended. But I see that youu put a
new optin Number value which is currently disabled. Is this optin the
solution for the previous mentioned problem?
5. VAR module: include a constant check box status is remembered by Gretl
but include a trend is allways unchecked when re-estimateing VAR. It is a
big possibility to forget to re-check this option.
What do you think about this?
From: Allin Cottrell <cottrell(a)wfu.edu>
Reply-To: Gretl list <gretl-users(a)ricardo.ecn.wfu.edu>
To: gretl-users(a)ricardo.ecn.wfu.edu
Subject: [Gretl-users] New Windows snapshot available
Date: Wed, 25 Jan 2006 16:14:51 -0500 (EST)
OK, there's now a snapshot with some of the new features we've been
discussing. It has matrix methods, 15-character variable names, and
a new lag-selection dialog box.
This version still needs to have the bugs shaken out of it, so I wouldn't
recommend feeding it to students yet. It is available at this URL only:
http://ricardo.ecn.wfu.edu/pub/gretl/gretl_install.exe
--
Allin Cottrell
Department of Economics
Wake Forest University, NC
_______________________________________________
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
10:18 p.m.
New subject: New Windows snapshot - PERFECT!
On Thu, 26 Jan 2006, john w wrote:
3. Pleeeaaassee put AIC and BIC tests for lag selection in VAR and
VECM. For
example in Tests module of VAR and VECM.
I've been thinking about this. First of all, note that you ALREADY get
those statistics in the VAR output. In case you haven't noticed, they're
right at the beginning.
But I assume that what you want is those criteria reported for several lag
orders, so you can choose the "right" lag order. The trouble is, I don't
think you want to check the information criteria AFTER having estimated
the model. The ideal thing to do, in my view, would be to have a separate
command (say, "lagselect") to give you an idea of the order of the VAR you
want to estimate. VECMs wouldn't need a separate procedure, since a VECM
is nothing but a VAR with a peculiar type of restriction on the
parameters.
The most linear way to accomplish this in a script would be something
like:
list VARvars = x y z
n = lagselect(VARvars,BIC)
var n VARvars
having the lagselect command issue some sort of output to let you check
what's happening. Note, by the way, that something like the above can be
achieved by some simple modifications to the script I sent a few days ago.
Now, how would you translate this into a GUI procedure? Maybe you can
select some variables from the main window and then (after right-clicking
or going through the menu) you get the output from lagselect, after which
you decide which lag structure you want and proceed with estimation. Or
would you rather have the option (in the VAR dialog) of automatically
selecting the lag order via AIC or BIC instead of specifying it manually?
Or both things?
If we reach a consensus on
a) whether this is necessary at all
b) if a), if it is urgently needed
c) if a) & b), the "right" way to do it,
then we can get something done.
Riccardo `Jack' Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
jack(a)dea.unian.it
http://www.econ.univpm.it/lucchetti
8:47 a.m.
New subject: New Windows snapshot - PERFECT!
El Jueves, 26 de Enero de 2006 23:18, jack escribió:
...The ideal thing to do, in my view, would be to have a separate
command (say, "lagselect") to give you an idea of the order of the VAR you
want to estimate. VECMs wouldn't need a separate procedure, since a VECM
is nothing but a VAR with a peculiar type of restriction on the
parameters.
The most linear way to accomplish this in a script would be something
like:
list VARvars = x y z
n = lagselect(VARvars,BIC)
var n VARvars
having the lagselect command issue some sort of output to let you check
what's happening. Note, by the way, that something like the above can be
achieved by some simple modifications to the script I sent a few days ago.
Now, how would you translate this into a GUI procedure? Maybe you can
select some variables from the main window and then (after right-clicking
or going through the menu) you get the output from lagselect, after which
you decide which lag structure you want and proceed with estimation. Or
would you rather have the option (in the VAR dialog) of automatically
selecting the lag order via AIC or BIC instead of specifying it manually?
Or both things?
If we reach a consensus on
a) whether this is necessary at all
b) if a), if it is urgently needed
c) if a) & b), the "right" way to do it,
then we can get something done.
There is a lag selection procedure in RATS (LAGSELEC.SRC), you can download it
from http://www.estima.com/procs_alpha.shtml
It works only for univariate series but I suppose this may give us some
indications about doing what you are thinking in gretl.
--
Ignacio Díaz-Emparanza
Dpto. de Economía Aplicada III (Econometría y Estadística)
UPV-EHU
7:06 p.m.
New subject: New Windows snapshot - PERFECT!
Yes I have noticed them in VAR output.
Yes, I'm pointing on: criteria reported for several lag orders, so you can
choose the "right" lag order.
They are a nice way to start building VARs.
You said:
The trouble is, I don't think you want to check the information criteria
AFTER having estimated
the model. The ideal thing to do, in my view, would be to have a
separate
command (say, "lagselect") to give you an idea of the order of the VAR you
want to estimate. VECMs wouldn't need a separate procedure, since a VECM
is nothing but a VAR with a peculiar type of restriction on the
parameters.
As you said, the best way will be to put them in VAR and VECM module.
You said:
Or would you rather have the option (in the VAR dialog) of automatically
selecting the lag order via AIC or BIC instead of specifying it
manually?
Or both things?
I agree to have both options (automatically and manually).
If this is not a problem it will be nice to have it. Firstly for students,
as when you start with learning VARs this is one of the first lessons and
the second thing is the reason that all professional software include this
option. I will be very happy that Gretl becomes a pro software-it has the
potential to be that!
Thnx.
>From: jack <jack(a)deanovell.unian.it>
>Reply-To: jack(a)deanovell.unian.it, Gretl list
><gretl-users(a)ricardo.ecn.wfu.edu>
>To: Gretl list <gretl-users(a)ricardo.ecn.wfu.edu>
>Subject: Re: [Gretl-users] New Windows snapshot - PERFECT!
>Date: Thu, 26 Jan 2006 23:18:56 +0100 (CET)
>
>On Thu, 26 Jan 2006, john w wrote:
>
> > 3. Pleeeaaassee put AIC and BIC tests for lag selection in VAR and VECM.
>For
> > example in Tests module of VAR and VECM.
>
>I've been thinking about this. First of all, note that you ALREADY get
>those statistics in the VAR output. In case you haven't noticed, they're
>right at the beginning.
>
>But I assume that what you want is those criteria reported for several lag
>orders, so you can choose the "right" lag order. The trouble is, I
don't
>think you want to check the information criteria AFTER having estimated
the model. The ideal thing to do, in my view, would be to have a
separate
command (say, "lagselect") to give you an idea of the order of the VAR you
want to estimate. VECMs wouldn't need a separate procedure, since a VECM
is nothing but a VAR with a peculiar type of restriction on the
parameters.
>
>The most linear way to accomplish this in a script would be something
>like:
>
> list VARvars = x y z
> n = lagselect(VARvars,BIC)
>
> var n VARvars
>
>having the lagselect command issue some sort of output to let you check
>what's happening. Note, by the way, that something like the above can be
>achieved by some simple modifications to the script I sent a few days ago.
>
>Now, how would you translate this into a GUI procedure? Maybe you can
>select some variables from the main window and then (after right-clicking
>or going through the menu) you get the output from lagselect, after which
>you decide which lag structure you want and proceed with estimation. Or
>would you rather have the option (in the VAR dialog) of automatically
selecting the lag order via AIC or BIC instead of specifying it
manually?
Or both things?
>
>If we reach a consensus on
>a) whether this is necessary at all
>b) if a), if it is urgently needed
>c) if a) & b), the "right" way to do it,
>then we can get something done.
>
>
>Riccardo `Jack' Lucchetti
>Dipartimento di Economia
>Università Politecnica delle Marche
>
>jack(a)dea.unian.it
>http://www.econ.univpm.it/lucchetti
>
>
>
>_______________________________________________
>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
3:24 a.m.
New subject: New Windows snapshot - PERFECT!
On Thu, 26 Jan 2006, john w wrote:
All I can say Allin: it's PERFECT!
Glad you like it! Actually I'm working a bit more on the design of
the lags dialog (how it relates to the main model dialog).
Thanks also for suggestions 1 to 3, which I will consider.
4. In VAR roots plot output: it is good have numerical values of
the roots in this plot but when some root is on or outside the
circle you can't see root values because plot window can't be
extended. But I see that youu put a new optin Number value which
is currently disabled. Is this optin the solution for the previous
mentioned problem?
Yes, it is the solution. What exactly do you mean when you say it's
"disabled"? I haven't tested the Windows version in particular but
it works fine here on Linux.
5. VAR module: include a constant check box status is remembered
by Gretl but include a trend is allways unchecked when
re-estimateing VAR. It is a big possibility to forget to re-check
this option.
"Include a constant" is not actually remembered, it's just the
default. If you uncheck it for one VAR, you'll have to uncheck it
again for the next. "Include a trend" is not the default, so it's
always off until checked. It would be easy enough to remember the
setting, but I wonder if that's really a good idea. It seems to me
that a sloppy user is just as likely
(a) to omit a trend by mistake, somehow assuming that gretl would
add it since the box was checked for a previous VAR, as
(b) to include a trend by mistake, failing to notice that the trend
box had remembered the choice to include a trend in a previous VAR.
Allin Cottrell
7:30 p.m.
New subject: New Windows snapshot - PERFECT!
Yes, I really like it!
And now it's really proffessional.
You said:
Yes, it is the solution. What exactly do you mean when you say
it's
"disabled"? I haven't tested the Windows version in particular but it
works fine here on Linux.
I use Gretl 1.5.0 Windows edition and it's not working. It's not disabled
but the error message comes out when clicking on that option. Something with
data error.
You said:
"Include a constant" is not actually remembered, it's
just the default. If
you uncheck it for one VAR, you'll have to uncheck it again for the next.
"Include a trend" is not the default, so it's always off until checked. It
would be easy enough to remember the setting, but I wonder if that's really
a good idea.
I think that it will be a good idea to remember the settings for trend too
just like as for the sesonals. They are remembered when re-estimateing the
same model. But I agree to reset them (not for a constant) when reopening
Gretl.
What do you think?
Also as a user I have some new suggestions (plus that from yesterday post):
1. It will be nice to include Cointegration vector (EC) graph and EC vector
data in the VECM module. Graph option under "Graphs module" and under "Add
to data set" EC graph values.
Now I can find the graph on the main Gretl console but it's a little bit non
practical solution.
2. In VECM output you can also put alphas wright after the Cointegratin
vectors. I know that you can see them in equatations (and this should stay)
but it's much clearer to put the alpha results once again after the vectors.
What do you say?
>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] New Windows snapshot - PERFECT!
>Date: Thu, 26 Jan 2006 22:24:24 -0500 (EST)
>
>On Thu, 26 Jan 2006, john w wrote:
>
>>All I can say Allin: it's PERFECT!
>
>Glad you like it! Actually I'm working a bit more on the design of the
>lags dialog (how it relates to the main model dialog).
>
>Thanks also for suggestions 1 to 3, which I will consider.
>
>>4. In VAR roots plot output: it is good have numerical values of the roots
>>in this plot but when some root is on or outside the circle you can't see
>>root values because plot window can't be extended. But I see that youu put
>>a new optin Number value which is currently disabled. Is this optin the
>>solution for the previous mentioned problem?
>
Yes, it is the solution. What exactly do you mean when you say
it's
"disabled"? I haven't tested the Windows version in particular but it
works fine here on Linux.
>
>>5. VAR module: include a constant check box status is remembered by Gretl
>>but include a trend is allways unchecked when re-estimateing VAR. It is a
>>big possibility to forget to re-check this option.
>
>"Include a constant" is not actually remembered, it's just the default.
If
>you uncheck it for one VAR, you'll have to uncheck it again for the next.
>"Include a trend" is not the default, so it's always off until checked.
It
>would be easy enough to remember the setting, but I wonder if that's really
>a good idea. It seems to me that a sloppy user is just as likely
>
>(a) to omit a trend by mistake, somehow assuming that gretl would add it
>since the box was checked for a previous VAR, as
>
>(b) to include a trend by mistake, failing to notice that the trend box had
>remembered the choice to include a trend in a previous VAR.
>
>Allin Cottrell
>
>
>_______________________________________________
>Gretl-users mailing list
>Gretl-users(a)ricardo.ecn.wfu.edu
>http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users
_________________________________________________________________
Be the first to hear what's new at MSN - sign up to our free newsletters!
http://www.msn.co.uk/newsletters
5:12 p.m.
New subject: New Windows snapshot - Again....
......perfect!
Thnx Allin and thnx Jack again. It's beautifull.
Windows version is working well.
I have some suggestions regardind this version:
1. Now constant, trend and seasonals are "remembered", but if you remoove
some variable from a list in VAR and VECM module this is not "remembered".
When re-estimateing again remooved variables are present.
It will be better that Gretl "remembers" only the LAST estimated
specification i.e. if some variable is remooved or unchecked then in the
next re-estimation this specifications must be "remembered".
2. It will be better to put "Lags" settings option wright after "Add"
option
in OLS module. It is more practical and the same logic is present in VAR and
VECM module.
Also try to consider the following suggestions:
1. Now what is missing is the F test of zero restrictions for exogenous
variables in VAR/VECM for every equatation and for a system as a whole !!!
It will be a great help.
2. It will be nice to include Cointegration vector (EC) graph and EC vector
data in the VECM module. Graph option under "Graphs module" and under "Add
to data set" EC graph values.
Now I can find the graph on the main Gretl console but it's a little bit non
practical solution.
2. In VECM output you can also put alphas wright after the Cointegratin
vectors. I know that you can see them in equatations (and this should stay)
but it's much clearer to put the alpha results once again after the vectors.
and
3. ...a recursive VAR/VECM estimation.
I think that this 3 suggestions will mainly complete VAR and VECM modules.
Thank you
_________________________________________________________________
Are you using the latest version of MSN Messenger? Download MSN Messenger
7.5 today! http://messenger.msn.co.uk
5:30 p.m.
New subject: New Windows snapshot - Again....
On Tue, 31 Jan 2006, john w wrote:
1. Now constant, trend and seasonals are "remembered", but
if you
remoove some variable from a list in VAR and VECM module this is
not "remembered". When re-estimateing again remooved variables are
present. It will be better that Gretl "remembers" only the LAST
estimated specification
I thought that was what gretl did. Could you give a specific
"recipe" for getting the behavior you describe (i.e. remove variable
how, from which list box)? Thanks.
2. It will be better to put "Lags" settings option wright
after
"Add" option in OLS module. It is more practical and the same
logic is present in VAR and VECM module.
I think that may be better. It raises the question of how to handle
IV estimation ("two-stage least squares") -- where should be button
be placed in that case?
Also try to consider the following suggestions:
1. Now what is missing is the F test of zero restrictions for
exogenous variables in VAR/VECM for every equatation and for a
system as a whole !!! It will be a great help.
I take it you mean exogenous variable other than deterministic ones
such as constant and trend?
I'll get back to your VECM suggestions shortly.
Allin.
5:55 p.m.
New subject: New Windows snapshot - Again....
Regarding 1.
I tried to estimate VAR with endogenous, deterministics and exogenous
variables. After taht I decided to re-estimate the model again. I again
clicked on VAR module. Now I remooved exogenous variables form the exogenous
list and clicked Ok. The model was estimated ok. Now I decided to again
re-estimate my model. I clicked again on VAR module. Exogenous variables
were present and they shouldn't be because I remooved them in the second
re-estimation. So, the last estimation wasn't remembered (the first was).
Regarding 2.
Yes, you are wright. Maybe the solution is to put "Lags" in OLS module after
Add option and to put it down in 2SLS next to Robust std errors box for
example.
Regarding F test on zero restrictions.
Hmm..it will be great to have F test of zero restrictions for deterministics
(const, trend and seasonals) and for exogenous variables. This will be
"Bulls eye"!
John
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] New Windows snapshot - Again....
Date: Tue, 31 Jan 2006 12:30:29 -0500 (EST)
On Tue, 31 Jan 2006, john w wrote:
>1. Now constant, trend and seasonals are "remembered", but if you remoove
>some variable from a list in VAR and VECM module this is not "remembered".
>When re-estimateing again remooved variables are present. It will be
>better that Gretl "remembers" only the LAST estimated specification
I thought that was what gretl did. Could you give a specific "recipe" for
getting the behavior you describe (i.e. remove variable how, from which
list box)? Thanks.
>2. It will be better to put "Lags" settings option wright after
"Add"
>option in OLS module. It is more practical and the same logic is present
>in VAR and VECM module.
I think that may be better. It raises the question of how to handle IV
estimation ("two-stage least squares") -- where should be button be placed
in that case?
>Also try to consider the following suggestions:
>
>1. Now what is missing is the F test of zero restrictions for exogenous
>variables in VAR/VECM for every equatation and for a system as a whole !!!
>It will be a great help.
I take it you mean exogenous variable other than deterministic ones such as
constant and trend?
I'll get back to your VECM suggestions shortly.
Allin.
_______________________________________________
Gretl-users mailing list
Gretl-users(a)ricardo.ecn.wfu.edu
http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users
_________________________________________________________________
Be the first to hear what's new at MSN - sign up to our free newsletters!
http://www.msn.co.uk/newsletters
6:09 p.m.
New subject: New Windows snapshot - Again....
On Tue, 31 Jan 2006, john w wrote:
Regarding 1.
I tried to estimate VAR with endogenous, deterministics and exogenous
variables. After taht I decided to re-estimate the model again. I again
clicked on VAR module. Now I remooved exogenous variables form the exogenous
list and clicked Ok. The model was estimated ok. Now I decided to again
re-estimate my model. I clicked again on VAR module. Exogenous variables were
present and they shouldn't be because I remooved them in the second
re-estimation.
OK, that's not what I saw on a quick test. I'll try some more
checking.
Regarding 2.
Yes, you are wright. Maybe the solution is to put "Lags" in OLS module after
Add option and to put it down in 2SLS next to Robust std errors box for
example.
OK, that makes sense.
Regarding F test on zero restrictions. Hmm..it will be great to
have F test of zero restrictions for deterministics (const, trend
and seasonals) and for exogenous variables. This will be "Bulls
eye"!
I think it would be best to implement this via the "omit" command
that's currently available for single-equation models -- that is,
extend it to cover VARs. In the GUI program this command would be
accessed via a menu item, "omit variables", under the Tests menu in
the VAR window (it's there at present for single equation models).
In the VAR case the variables displayed as candidates for omission
would be the deterministic/exogenous terms. Then the user can
choose which subset to test.
Allin.
6:27 p.m.
New subject: New Windows snapshot - Again....
I use Windows version if this can help you.
Yes, that's it Allin!
"Omit" variables is the solution. But why not to extend it on endogenous
variables too!? (Plus deterministics and exogenous variables). This will be
just perfect!
Also, I think that this is possible for VECMs too.
As we are talking about causalities, one of the possible feature could be
the possibility to test Granger causality (block and bivariate or pairwise).
What do you think?
John
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] New Windows snapshot - Again....
Date: Tue, 31 Jan 2006 13:09:09 -0500 (EST)
On Tue, 31 Jan 2006, john w wrote:
>Regarding F test on zero restrictions. Hmm..it will be great to have F
>test of zero restrictions for deterministics (const, trend and seasonals)
>and for exogenous variables. This will be "Bulls eye"!
I think it would be best to implement this via the "omit" command that's
currently available for single-equation models -- that is, extend it to
cover VARs. In the GUI program this command would be accessed via a menu
item, "omit variables", under the Tests menu in the VAR window (it's there
at present for single equation models). In the VAR case the variables
displayed as candidates for omission would be the deterministic/exogenous
terms. Then the user can choose which subset to test.
Allin.
_______________________________________________
Gretl-users mailing list
Gretl-users(a)ricardo.ecn.wfu.edu
http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users
_________________________________________________________________
The new MSN Search Toolbar now includes Desktop search!
http://toolbar.msn.co.uk/
6:44 p.m.
New subject: New Windows snapshot - Again....
On Tue, January 31, 2006 19:27, john w wrote:
I use Windows version if this can help you.
Yes, that's it Allin!
"Omit" variables is the solution. But why not to extend it on endogenous
variables too!? (Plus deterministics and exogenous variables). This will be
just perfect!
Hmm.. generalising here is tricky. If you omit SOME lags from a VAR, then you
lose the justification for estimating it via OLS, because if the regressors
vary across equations you should use SUR (that gretl provides, by the way)
instead. So I think endogenous variables should be left alone.
Also, I think that this is possible for VECMs too.
As we are talking about causalities, one of the possible feature could be
the possibility to test Granger causality (block and bivariate or pairwise).
What do you think?
You've already got that: the F-tests under the individual regressions are
precisely Granger-causality tests.
--
Riccardo "Jack" Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
7:19 p.m.
New subject: New Windows snapshot - Again....
Hmm...I saw omitting endogenous variables in PcGive. It is possibile. These
are so called Exclusion restrictions and all variables in VAR can be tested.
You can test all the variables in VAR using F test but Exclusion
restrictions are done via Subset Chi^2 (square) test and you can test
endogenous, deterministics and exogenous variables.
Are Granger tests doable for VECMs ? As tere are no F tests in VECM
output. (my question excludes testing beta)
In previous posts I was talking about VAR/VECM confidence intervals. They
are set as default in plot console to be errorbars and this can't be
changed. Is it possible to activate this option so user can set it freely
for example to be lines?
Thanx
John
From: "Riccardo Jack" Lucchetti""
<r.lucchetti(a)univpm.it>
Reply-To: r.lucchetti(a)univpm.it, Gretl list
<gretl-users(a)ricardo.ecn.wfu.edu>
To: "Gretl list" <gretl-users(a)ricardo.ecn.wfu.edu>
Subject: Re: [Gretl-users] New Windows snapshot - Again....
Date: Tue, 31 Jan 2006 19:44:01 +0100 (CET)
On Tue, January 31, 2006 19:27, john w wrote:
> I use Windows version if this can help you.
>
> Yes, that's it Allin!
> "Omit" variables is the solution. But why not to extend it on endogenous
> variables too!? (Plus deterministics and exogenous variables). This will
be
> just perfect!
>
Hmm.. generalising here is tricky. If you omit SOME lags from a VAR, then
you
lose the justification for estimating it via OLS, because if the regressors
vary across equations you should use SUR (that gretl provides, by the way)
instead. So I think endogenous variables should be left alone.
> Also, I think that this is possible for VECMs too.
>
> As we are talking about causalities, one of the possible feature could
be
> the possibility to test Granger causality (block and bivariate or
pairwise).
> What do you think?
You've already got that: the F-tests under the individual regressions are
precisely Granger-causality tests.
--
Riccardo "Jack" Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
_______________________________________________
Gretl-users mailing list
Gretl-users(a)ricardo.ecn.wfu.edu
http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users
_________________________________________________________________
Be the first to hear what's new at MSN - sign up to our free newsletters!
http://www.msn.co.uk/newsletters
8:15 p.m.
New subject: Clear function
I have one question regarding "Clear" function for VAR/VECM module.
I noticed that "Clear" function clears only endogenous variables while
exogenous not.
It should clear exogenous too.
Basicly, this function clears all except const, trend, seasonals, and number
of lags? Wright?
Maybe it will be better to be something like "reset" function i.e. clears
all endogenous and exogenous variables, returns lags to a max value of 12,
clears trend and seasonals and leaves constant checked. In other words,
clears everything except constant. This is for VAR module.
And for VECM module the same thing except - returns number of coint rank to
1, and leaves "Unrestricted constant" checked.
What do you say?
Sorry for so much questions.....but thanx to you Gretl will be the best
software (this includes paid software too).
God bless you all guys.
Be sure that I will mention you in my works.
_________________________________________________________________
The new MSN Search Toolbar now includes Desktop search!
http://toolbar.msn.co.uk/
8:25 p.m.
New subject: Clear function
On Tue, 31 Jan 2006, john w wrote:
I have one question regarding "Clear" function for VAR/VECM
module.
I noticed that "Clear" function clears only endogenous variables while
exogenous not.
It should clear exogenous too.
Basicly, this function clears all except const, trend, seasonals, and number
of lags? Wright?
The "Clear" function needs updating: it hasn't been revised since
a time when the model dialog box was a good deal simpler. Your
suggestions are good.
Allin.
8:39 p.m.
New subject: Clear function
Thank you Allin.
Will you take my previous requests in count ("omit" function and confidence
intervals plot ) ?
_________________________________________________________________
Be the first to hear what's new at MSN - sign up to our free newsletters!
http://www.msn.co.uk/newsletters
10:26 a.m.
New subject: Clear function
Thnx Allin.
I was thinking about "Lags" option in OLS and 2SLS. Maybe the best thing
will be to put the option next to "Robust std error" option. For both, OLS
(not below "Add" option as we said before) and 2SLS.
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] Clear function
Date: Tue, 31 Jan 2006 16:11:36 -0500 (EST)
On Tue, 31 Jan 2006, john w wrote:
>Will you take my previous requests in count ("omit" function and
>confidence intervals plot ) ?
Yes, at least up to a point ;-)
Allin
_______________________________________________
Gretl-users mailing list
Gretl-users(a)ricardo.ecn.wfu.edu
http://ricardo.ecn.wfu.edu/mailman/listinfo/gretl-users
_________________________________________________________________
The new MSN Search Toolbar now includes Desktop search!
http://toolbar.msn.co.uk/
9:07 p.m.
New subject: New Windows snapshot - Again....
On Tue, January 31, 2006 20:19, john w wrote:
Hmm...I saw omitting endogenous variables in PcGive. It is possibile.
These
are so called Exclusion restrictions and all variables in VAR can be tested.
You can test all the variables in VAR using F test but Exclusion
restrictions are done via Subset Chi^2 (square) test and you can test
endogenous, deterministics and exogenous variables.
Of course it is possible. VARs are rather trivial parametric models for which
standard inference methods apply (at least in the stationary case --- see
below). My point was, general exclusion tests are often pointless in VARs,
since these models don't aim at a structural representation of the data, but
only at encapsulating conveniently their time-series properties. If you feel
like carving out variables and/or lags, you're much better off with a
simultaneous equations system.
Are Granger tests doable for VECMs ? As tere are no F tests in
VECM
output. (my question excludes testing beta)
This brings back a point you raised some time ago (see
http://ricardo.ecn.wfu.edu/pipermail/gretl-users/2006-January/000326.html and
subsequent messages). I had a look at JMulTi. They do Granger causality tests
in cointegrated systems by using a very ingenious method, proposed by
Lütkepohl (usurprisingly) and Dolado, that bypasses most of the inferential
difficulties by estimating a levels VAR of higher order than necessary. It's
not difficult to do the same in a script, I think (but I have to dig a little
deeper into this).
--
Riccardo "Jack" Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona
10:51 a.m.
New subject: New Windows snapshot - Again....
Hi Jack,
Of course it is possible. VARs are rather trivial parametric models
for
which
standard inference methods apply (at least in the stationary case --- see
below). My point was, general exclusion tests are often pointless in VARs,
since these models don't aim at a structural representation of the data,
but
only at encapsulating conveniently their time-series properties. If you
feel
like carving out variables and/or lags, you're much better off with a
simultaneous equations system.
Yes, I remember. we were already talking about this. And for now the best
thing will be to leave F tests of zero restrictions for endobenous variables
as they are now but to add "Omit" option for deterministics, seasonals and
exogenous variables.
There is a simillar option in JMulTi. It's an LR (likelihood ratio) test. In
"Subset restrictions" pannel you can impose zero restrictions on
coefficients before estimating the model.
> Are Granger tests doable for VECMs ? As tere are no F tests
in VECM
> output. (my question excludes testing beta)
This brings back a point you raised some time ago (see
http://ricardo.ecn.wfu.edu/pipermail/gretl-users/2006-January/000326.html
and
subsequent messages). I had a look at JMulTi. They do Granger causality
tests
in cointegrated systems by using a very ingenious method, proposed by
Lütkepohl (usurprisingly) and Dolado, that bypasses most of the inferential
difficulties by estimating a levels VAR of higher order than necessary.
It's
not difficult to do the same in a script, I think (but I have to dig a
little
deeper into this).
Ok. If it's possible it will be great to have it.
If you remember we were talking about Forrecast error variance decompositin
graph and you said that it's possible to produce it.
Ok. For example, FEVD for variable X can be produced by lines (in time) of
Y,Z and T variables. On x-axis could be periods(horizon) and on y-axis
decompositions (up to a max number of 100).
What do you say?
_________________________________________________________________
Are you using the latest version of MSN Messenger? Download MSN Messenger
7.5 today! http://messenger.msn.co.uk
8:10 p.m.
New subject: Cointegration test - Help
Hi guys,
got one question for you regarding coint test - Johansen.
Determing the number of coint vectors (CV) is the first step of VECM
(besides testing for a unit root).
This means that for determing the wright number of CVs you must estimate a
system which don't suffer of serial correlation, no arch effects,
non-normality etc. The results and estimates from cointegration tests are
then the base for VECM.
My question is: How you choose the wright model i.e. the wright
specification for coint test in Gretl?
Do you look at individual equatations tests in Gretl output cointegration
test?
Is there any additional command to test system as a whole?
Do you start with AIC and BIC var lag order tests? But they do not guarantee
that the model is miss-specified (no serial correlation etc.)
In my question cointegration cases are ignored.
Thnx
_________________________________________________________________
The new MSN Search Toolbar now includes Desktop search!
http://toolbar.msn.co.uk/
4:33 p.m.
New subject: gretl and (old) gtk-1.2
I wonder, are any list members using a recent version of gretl built
with gtk-1.2? Or does anyone know of a colleague who's doing so?
If you're running gretl on Windows you're using gtk-2.0. You'd be
using gtk-1.2 only if (a) you've downloaded one of
gretl-VERSION-1rh73.i386.rpm
gretl-VERSION-1generic.i386.rpm
or (b) you compiled gretl yourself on a gtk-1.2 system. You can
check that by doing
gretl --dump
at the command prompt to dump gretl's configuration. The gtk
version should be stated at the top of the output file.
The reason I ask is this: gtk-1.2 was been officially "obsolete" for
years now, but to date we have continued to support it. It would
clean up the GUI code considerably if we could drop support for
gtk-1.2 at this point.
Allin Cottrell
6898
days inactive
6905
days old
22 comments
5 participants
participants (5)
-
Allin Cottrell -
Ignacio Díaz-Emparanza -
jack -
john w -
Riccardo Jack Lucchetti