BACE Estimation difficulties
by Brian Revell
I thought i would try this option estimating a simple model (always the
best way to gain experience and confidence with the routine).
Independent variable, 1 indep variable . No lag on Y. . Zero selected for
out of sample forecast Constant -can be omitted. Best 4 models selected
-again as the default.
This was the output:-
Dependent viariable was removed from the X list.* It is not there -there is
only a single indep variable. *
Error message from BACE_GUI():
Average model size should be greater then 0 and lower than K !!! *Average
model size specified is the default 1*
*Some guidance wlecomed please*
*Brian*
4 months, 1 week
BACE GUI Output
by Brian Revell
Just a minor point. It would be helpful if the sample size or sample time
period was displayed in BACE outputfrom the GUI.. No need to print it out
with each model iteration -with the initial OLS model would suffice.
Brian
4 months, 2 weeks
Combined irf in VAR/VEC model.
by Olasehinde Timmy
Dear Admin,
I am trying to plot the impulse response(irf) after estimating a 5-variable
VAR/VEC model, however, I realized that the combined irf graphics option
disappeared. But when I reduced the variables to four, it reappeared.
Was this a bug or something? Then I would like to know how to go about it.
Regards
Timmy.
4 months, 2 weeks
Re: How to apply the HAC option in Gretl?
by Cottrell, Allin
On Mon, Nov 7, 2022 at 12:29 PM Torbjørn Lorentzen
<torbjorn.lorentzen(a)outlook.com> wrote:
>
> I'm a new user of Gretl. I'm using ols in the estimation of a time series model. T want to us ethe HAC-option to correct for autocorrelation and heteroscedasticity. I tried the robust option, but to what I can see it only corrects for heteroscedasticity? Great if anyone could help me with the Gretl script for HAC and how to put it together with the ols command. All suggestions are appreciated 😊
Hi Tobben. If you go to the menu item /Tools/Preferences/General
you'll see an "HCCME" tab. That lets you set your preferences for
robust covariance matrix estimation. Note that the HAC option is only
available when your dataset is recognized as time-series.
In scripting usage the relevant option is --robust. What exactly that
does depends on the aforementioned setting. But if the data are time
series gretl will use the HAC estimator by default. In that case you
should see a line like the following in the estimation output:
HAC standard errors, bandwidth 4 (Bartlett kernel)
Allin Cottrell
4 months, 3 weeks
Some suggestions on the "Number of cases ’correctly predicted'" from a logit/probit outputs
by Fred Engst
Hi Jack, Allin, Sven and all others on the gretl team,
As you know, "Number of cases ’correctly predicted’" in a logit/probit model can be miss-leading even in a 50/50 split case.
What we should be comparing is not zero ‘correctly predicted’, but rather random assignments based on sample mean.
If a sample is 50/50 split, a random assignment would get 50% "correctly predicted", in theory. If our model's 'correctly predicted' is 70%, we are only 20 percentage points higher than a model based on random assignment, representing an improvement over the random assignment model by only 40%.
Thus, I would like to propose an alternative output from gretl, i.e. the “Extra number of cases 'correctly predicted' over random assignment” (or something like that), call this dot_R-square perhaps.
Dot_R-saure = (Y_hat_model - Y_hat_random) / (1-Y_hat_random)
where Y_hat_model = sum(Y_hat_model_i=Y_i)/N
Y_hat_random = Y_hat^2 + (1-Y_hat)^2
Y_hat is the sample mean
Y_hat_model_i = Pro(Y_i = 1) >Y_hat
Pro(Y_i = 1) >Y_hat = 1, if Pro(Y_i = 1) >Y_hat is true
Unlike the McFadden R-squared, the interpretation of this is fairly straight forward, i.e. the percent that our model is better off than a model based on random assignment.
Best,
Fred
4 months, 3 weeks
deseas function arima option
by Javier Sansa
I have been trying to use the different options of the deseas function in
gretl. All the options I used seem to work as intended except the arima
option. To be more specific I have tried to build a deseasonalizing
procedure in which a certain arima model is imposed. According to the gretl
reference manual this should be done by building a bundle and using "arima"
and a 6x1 matrix which defines the model, like in the following example:
matrix start ={0,2,1,0,1,1}
bundle b=defbundle(verbose=2,seats=1,outliers=7, arima=start)
The bundle definition works, i.e a bundle b is built, but function deseas
fails because it does not recognize the arima key
I attach a file with data and a short script should something be unclear.
Please, let me know if I failed to follow the manual or there is something
strange going on with the program.
And since we are at the end of 2022, a happy new year!
Javier Sansa
4 months, 3 weeks
count panel data - poisson model with dummies?
by lars.ahnland＠outlook.com
Hi,
I have count data with what appears to be poisson distribution of the dependent variable in panel data. There is no automatic method for doing this in Gretl as I have understood it, so I am using a count, poisson, model with cross-sectional dummies. Gretl deletes the last dummy so it becomes a fixed effects model, so intuitively, it looks as if the model should be ok. Do you agree? One problem is that Gretl fails to include an overdispersion (Chi-square) test, so I am not sure the poisson model is valid. Can you help me with this too?
Best regards,
Lars Ahnöand
4 months, 3 weeks
Function package DHF_test is going to be retired
by Sven Schreiber
Hello everybody,
this is to let you know that the contributed function package "DHF_test"
is going to be removed from the public package server soon. If you rely
on that package for whatever reason, you can keep a local copy on your
system(s). (And also perhaps write us about the fact that you're using
it, which would be interesting as well.)
The package DHF_test implements an early test for seasonal unit roots,
its name stands for Dickey-Hasza-Fuller. But for seasonal unit root
testing we currently recommend to use the "GHegy" package instead, which
also has been available for many years on the gretl package server, and
which refers to the HEGY publication (Hylleberg/Engle/Granger/Yoo, 1990).
DHF_test package author Marcin obviously knows about this
reorganization, as he is also a member of the gretl team.
If you have any remaining questions about this (and perhaps also about
seasonal unit root testing in gretl), please go ahead and ask them.
cheers
sven
4 months, 3 weeks
drag-and-drop a datafile now also supports appending
by Sven Schreiber
Hi everybody,
here's some info on a brand new little interface feature that you might
not notice otherwise. (It's in the development snapshots already.)
Many of you probably know already that you can drag-and-drop a datafile
with the mouse onto the gretl main window to open it. (gretl formats
.gdt or .gdtb, and actually other supported formats as well. If you have
the gretl console set to be merged into the main window on the
right-hand side, this only works in the left area of the window where
the variables' listing appears.)
Now when you have a dataset already active and drag-and-drop another
file onto gretl, appending the new datafile contents to the old dataset
is also possible. Of course, in such a situation gretl cannot know by
itself what exactly you want, so a dialog window will pop up asking you
for your choice: open as new dataset or append to the old one?
The type of action directly corresponds to gretl's commands 'append' and
'join'. As you may know, 'append' is easier but 'join' is more general
and powerful.
(Are animated GIFs actually supported on this mailing list? Maybe this
would be a good case to prepare and attach one, showing the dragging and
dropping....)
cheers
sven
4 months, 3 weeks
Contributed function packages in 2022
by Sven Schreiber
Dear all,
happy new year to everybody!
Now that the year 2022 is over, just a small piece of trivial
information: There are 48 contributed function packages that have been
updated during the last year - although it has to be said that in many
cases the updates consisted of smaller syntax adjustments.
This number doesn't include the so-called "addons": those are also
function packages written in the hansl scripting language (like the
contributed packages), but they are more tightly integrated into the
gretl infrastructure and are automatically shipped with each gretl
install. Currently there are 10 addons (see Help / look for addons),
half of which have been updated during 2022. For comparison, there are
155 contributed packages in total.
We would like to encourage everybody to get in the business of authoring
and contributing function packages. The scripting language hansl is easy
to learn, but powerful. And in general gretl offers documentation that
is quite comprehensive: See Help / A Hansl Primer for an introduction to
the language (in combination with the general gretl user guide, of
course), and Help / Gretl Function Packaging Guide.
Every package version has to go through a moderation process, where we
as the moderation team make sure that (a) the package actually runs and
works, and (b) has at least a minimal amount of documentation.
Especially for first-time contributors, we're happy to help with
problems and provide some hints. However, the package authors obviously
remain responsible for what their code does and what results it
produces. (Apart from the obvious fact that like any GPL'ed free / open
source software, neither gretl itself nor the contributed packages come
with any kind of warranties.)
cheers
sven
5 months
