3:06 a.m.
I am a new user to Gretl, and am using version 1.2.9. I have noticed
that frequently when I attempt to run an ARMA or GARCH model I get a
message "Convergence Criteria not met". I know what this means, but it
seems to happen frequently with data sets that do not experience this
problem when I use them in "R" or other packages. Is this a known
problem, or is there some parameter I need to adjust to make this go
away?
6:54 a.m.
On Mon, 25 Oct 2004, Blaise Roncagli wrote:
I am a new user to Gretl, and am using version 1.2.9. I have
noticed
that frequently when I attempt to run an ARMA or GARCH model I get a
message "Convergence Criteria not met". I know what this means, but it
seems to happen frequently with data sets that do not experience this
problem when I use them in "R" or other packages. Is this a known
problem, or is there some parameter I need to adjust to make this go
away?
This is a known issue for GARCH models, and some modifications that
somewhat alleviate the problem are already in the CVS version. For ARMA
models, this should happen far less frequently. Could you post an example?
Riccardo `Jack' Lucchetti
Dipartimento di Economia
Università di Ancona
jack(a)dea.unian.it
http://www.econ.unian.it/lucchetti
12:34 p.m.
On Mon, 25 Oct 2004, Blaise Roncagli wrote:
I am a new user to Gretl, and am using version 1.2.9. I have
noticed
that frequently when I attempt to run an ARMA or GARCH model I get a
message "Convergence Criteria not met". I know what this means, but it
seems to happen frequently with data sets that do not experience this
problem when I use them in "R" or other packages. Is this a known
problem, or is there some parameter I need to adjust to make this go
away?
It is a known problem with gretl's GARCH function, and I think it is
fixed for the forthcoming version 1.3.0. In the meantime you should
be able to work around the problem by defining a new variable that
is a scaled version of the original dependent variable (call it Y):
genr s = sd(Y)
genr Yscale = Y/s
then run GARCH on Yscale.
As for ARMA, I'm not so sure. Could you send me an example of a
data file that is processed OK by R but not gretl? That would be
helpful in troubleshooting. Thanks.
Allin Cottrell
2:30 a.m.
Thanks for the quick response. I have attached a dataset called
"sampledata.xls" that contains the following returns time series data:
NASDAQ, TBILL3 (3-month US treasury Bills). When running ARMA(1,1) and
GARCH(1,1) models on this data using Gretl 1.2.9 (build date 7/23/2004) on
the PC platform I get the following results:
NASDAQ: ARMA - no convergence, GARCH - runs OK
TBILL3: ARMA - no convergence, GARCH - no convergence
Note that I have not yet scaled these as you recommended, but I will try
that. I thought these might be good tests for your new release version,
which you say you have fixed. I have run these in R, and they do converge.
By the way, when will the new version 1.3.0 be released for the PC platform?
We are using Gretl in a graduate course in Finance and both the students and
Professor would be very interested in the new version. Right now, we use
Gretl whenever we can, but these convergence problems have forced us to use R
or Eviews for some of our assignments, though Gretl is the preferred tool. I
would be happy to relay to the class the information that a new release is
imminent.
Thanks for your time and assistance.
- Blaise Roncagli
Allin Cottrell wrote:
On Mon, 25 Oct 2004, Blaise Roncagli wrote:
> I am a new user to Gretl, and am using version 1.2.9. I have noticed
> that frequently when I attempt to run an ARMA or GARCH model I get a
> message "Convergence Criteria not met". I know what this means, but it
> seems to happen frequently with data sets that do not experience this
> problem when I use them in "R" or other packages. Is this a known
> problem, or is there some parameter I need to adjust to make this go
> away?
It is a known problem with gretl's GARCH function, and I think it is
fixed for the forthcoming version 1.3.0. In the meantime you should
be able to work around the problem by defining a new variable that
is a scaled version of the original dependent variable (call it Y):
genr s = sd(Y)
genr Yscale = Y/s
then run GARCH on Yscale.
As for ARMA, I'm not so sure. Could you send me an example of a
data file that is processed OK by R but not gretl? That would be
helpful in troubleshooting. Thanks.
Allin Cottrell
5:27 a.m.
On Tue, 26 Oct 2004, Blaise Roncagli wrote:
Thanks for the quick response. I have attached a dataset called
"sampledata.xls" that contains the following returns time series data:
NASDAQ, TBILL3 (3-month US treasury Bills). When running ARMA(1,1) and
GARCH(1,1) models on this data using Gretl 1.2.9 (build date 7/23/2004) on
the PC platform I get the following results:
NASDAQ: ARMA - no convergence, GARCH - runs OK
TBILL3: ARMA - no convergence, GARCH - no convergence
Thanks very much for the sample data. I wonder if perhaps there is
an issue of conversion of the data from Excel -- in particular, if
missing values are being handled correctly by gretl. I'm attaching
a version of the data in gretl format. My findings are
* ARMA(1,1) works OK for both series
* GARCH(1,1), using the current CVS code, works for NASDAQ, but not
for TBILL3
Looking at a time series plot of TBILL3, I'm not surprised that a
univariate GARCH routine has a hard time (in fact, I'm quite
surprised that R could produce estimates). The huge spike in the
T-bill rate in the early 1980s just doesn't look like a case of
conditional heteroskedasticity. But I'll try running R on the data
tomorrow.
For Windows, you can get a current snapshot, which is close to the
final 1.3.0 release, at
ftp://ricardo.ecn.wfu.edu/pub/gretl/gretl_install.exe
Allin Cottrell
7:22 p.m.
New subject: convergence of ARMA and GARCH models
On Tue, 26 Oct 2004, Blaise Roncagli wrote:
Thanks for the quick response. I have attached a dataset called
"sampledata.xls"...
I'm currently trying to do a comparison of GARCH results from gretl
and R (the ARMA results seem pretty much the same).
My problem at present is that I can't get R to produce any sensible
GARCH estimates. I'm using R 2.0.0 on Linux, with the "tseries"
library version 0.9-23. With Blaise's two series, and also with the
benchmark Bollerslev-Ghysels dataset, the GARCH "a1" and "b1"
estimates are printed as "0.050000", regardless of the particular
series used. (I have printed the series out from R to check that
the data are right.) The B-G results do not remotely resemble the
benchmark results.
I'm using the defaults (1,1), as in:
y.garch <- garch(y)
summary(y.garch)
Any ideas on what I might be doing wrong?
Allin Cottrell
9:10 a.m.
New subject: convergence of ARMA and GARCH models
On Wed, 27 Oct 2004, Allin Cottrell wrote:
On Tue, 26 Oct 2004, Blaise Roncagli wrote:
> Thanks for the quick response. I have attached a dataset called
> "sampledata.xls"...
I'm currently trying to do a comparison of GARCH results from gretl
and R (the ARMA results seem pretty much the same).
My problem at present is that I can't get R to produce any sensible
GARCH estimates. I'm using R 2.0.0 on Linux, with the "tseries"
library version 0.9-23. With Blaise's two series, and also with the
benchmark Bollerslev-Ghysels dataset, the GARCH "a1" and "b1"
estimates are printed as "0.050000", regardless of the particular
series used. (I have printed the series out from R to check that
the data are right.) The B-G results do not remotely resemble the
benchmark results.
Folks, I'm attaching a file that pretty much explains what the trouble
with TBILL is: for a Garch(1,1) you have 4 parameters:
y_t = k + e_t
v(e_t) = \omega + a e_{t-1}^2 + b v(e_{t-1})
Where v(.) = conditional variance.
I run a little experiment:
1) keep k fixed at sample mean of y
2) keep omega at (sample variance)*(1-a-b)
3) plot the likelihood for all admissible values of a and b (grid width =
0.1)
As you can see by the attached files, this is one of the rare (but not
exceptional) cases when the sample likelihood doesn't have a maximum in
the interior of the parameter space.
It must be said that the original series looks very much I(1), so applying
a garch model to the TBILL series is rather meaningless anyway.
Riccardo `Jack' Lucchetti
Dipartimento di Economia
Università di Ancona
jack(a)dea.unian.it
http://www.econ.unian.it/lucchetti
9:27 a.m.
The new build of Gretl did fix my problems. Thank you very much for your quick
response and your help! Now we can continue to use Gretl in my class without
having to resort to other packages. Regards.
- Blaise Roncagli
Allin Cottrell wrote:
On Tue, 26 Oct 2004, Blaise Roncagli wrote:
> Thanks for the quick response. I have attached a dataset called
> "sampledata.xls" that contains the following returns time series data:
> NASDAQ, TBILL3 (3-month US treasury Bills). When running ARMA(1,1) and
> GARCH(1,1) models on this data using Gretl 1.2.9 (build date 7/23/2004) on
> the PC platform I get the following results:
>
> NASDAQ: ARMA - no convergence, GARCH - runs OK
> TBILL3: ARMA - no convergence, GARCH - no convergence
Thanks very much for the sample data. I wonder if perhaps there is
an issue of conversion of the data from Excel -- in particular, if
missing values are being handled correctly by gretl. I'm attaching
a version of the data in gretl format. My findings are
* ARMA(1,1) works OK for both series
* GARCH(1,1), using the current CVS code, works for NASDAQ, but not
for TBILL3
Looking at a time series plot of TBILL3, I'm not surprised that a
univariate GARCH routine has a hard time (in fact, I'm quite
surprised that R could produce estimates). The huge spike in the
T-bill rate in the early 1980s just doesn't look like a case of
conditional heteroskedasticity. But I'll try running R on the data
tomorrow.
For Windows, you can get a current snapshot, which is close to the
final 1.3.0 release, at
ftp://ricardo.ecn.wfu.edu/pub/gretl/gretl_install.exe
Allin Cottrell
------------------------------------------------------------------------
Name: blaise.gdt
blaise.gdt Type: Plain Text (TEXT/PLAIN)
Encoding: BASE64
7112
days inactive
7115
days old
7 comments
4 participants
participants (4)
-
Allin Cottrell -
Blaise Roncagli -
jack -
jack