12:54 p.m.
Hello gretl users!
I want to report a bug. I saved a session with several VAR models and
graphs. When I try to re-open the session and to run an impulse-response
analysis without the bootstrapped confidence intervals there occurs no
problem. But running the impulse-response analysis with confidence
intervals lets gretl crash and gives this output:
artur@artur-laptop:~$ gretl
setlocale(LC_NUMERIC, "de_DE.UTF-8") returned de_DE.UTF-8
get_gretl_charset: using UTF-8
Lese Sitzungs-Datei /home/artur/Eigene_Daten/Uni/VWL/Unterricht/Wachstum
und Verteilung/Bachelorabschlussarbeit/Verteilung, Wachstum und
technologischer Fortschritt/Sitzung gretl.gretl
unzip_session_file: returning 0
session_name_from_session_file: /home/artur/Eigene_Daten/Uni/VWL/Unterricht/Wachstum und
Verteilung/Bachelorabschlussarbeit/Verteilung, Wachstum und technologischer
Fortschritt/Sitzung gretl.gretl -> Sitzung gretl
model file: fname='.Sitzung gretl/model.1', type=3
VAR_retrieve_equations: got 4 models (neqns = 4), err = 0
rebuild_session_model: IN_GUI_SESSION yes, IN_MODEL_TABLE no
IN_NAMED_STACK yes, IS_LAST_MODEL no
error 0 from session_append_model
model file: fname='.Sitzung gretl/model.2', type=3
VAR_retrieve_equations: got 3 models (neqns = 3), err = 0
rebuild_session_model: IN_GUI_SESSION yes, IN_MODEL_TABLE no
IN_NAMED_STACK yes, IS_LAST_MODEL no
error 0 from session_append_model
model file: fname='.Sitzung gretl/model.3', type=3
VAR_retrieve_equations: got 3 models (neqns = 3), err = 0
rebuild_session_model: IN_GUI_SESSION yes, IN_MODEL_TABLE no
IN_NAMED_STACK yes, IS_LAST_MODEL no
error 0 from session_append_model
model file: fname='.Sitzung gretl/model.4', type=3
VAR_retrieve_equations: got 3 models (neqns = 3), err = 0
rebuild_session_model: IN_GUI_SESSION yes, IN_MODEL_TABLE no
IN_NAMED_STACK yes, IS_LAST_MODEL no
error 0 from session_append_model
model file: fname='.Sitzung gretl/model.5', type=3
VAR_retrieve_equations: got 3 models (neqns = 3), err = 0
rebuild_session_model: IN_GUI_SESSION yes, IN_MODEL_TABLE no
IN_NAMED_STACK yes, IS_LAST_MODEL no
error 0 from session_append_model
model file: fname='.Sitzung gretl/model.6', type=3
VAR_retrieve_equations: got 3 models (neqns = 3), err = 0
rebuild_session_model: IN_GUI_SESSION yes, IN_MODEL_TABLE no
IN_NAMED_STACK yes, IS_LAST_MODEL no
error 0 from session_append_model
Lesen 134919 bytes von Daten...
Opened session datafile '.Sitzung gretl/tes mit namen.gdt'
Segmentation fault
artur@artur-laptop:~$
Best regards
Artur
8:47 p.m.
On Thu, 14 Aug 2008, artur wrote:
I want to report a bug. I saved a session with several VAR
models and graphs. When I try to re-open the session and to run
an impulse-response analysis without the bootstrapped confidence
intervals there occurs no problem. But running the
impulse-response analysis with confidence intervals lets gretl
crash...
Thanks for spotting this. Fixed in CVS and today's Windows
snapshot. We were not saving into the session file all the
information needed to run a bootstrapped IRF (and then the
crash was due to not checking whether the information was
really there or NULL).
Allin.
9:03 p.m.
New subject: Can gretl do the moving forward forecast?
Hi,
Can someone take a look to see how I can program in gretl to achieve below?
Below is what I wrote for one run. How to program it so that it can run in loop?
setobs 12 2000:01
smpl 2000:01 2005:12
ols y const x1 x2 x3
fcast 2006:01 2006:01 --out-of-sample
setobs 12 2000:02
smpl 2000:02 2006:01
ols y const x1 x2 x3
fcast 2006:02 2006:02 --out-of-sample
...
(Each time to move one month forward.)
Also how to store the series data generated in those runs in a way convenient for further
process?
Thank you for your help!
Regards,
Rebecca
1:36 a.m.
New subject: Can gretl do the moving forward forecast?
On Thu, 28 Aug 2008, Rebecca Zhang wrote:
Can someone take a look to see how I can program in gretl to
achieve below?
Yes, will do.
Below is what I wrote for one run. How to program it so that it
can run in loop?
setobs 12 2000:01
smpl 2000:01 2005:12
ols y const x1 x2 x3
fcast 2006:01 2006:01 --out-of-sample
setobs 12 2000:02
smpl 2000:02 2006:01
ols y const x1 x2 x3
fcast 2006:02 2006:02 --out-of-sample
...
(Each time to move one month forward.)
First of all, unless I'm missing something, I don't think that the
"setobs" command is what you want in your loop. This does not
advance the start of the sample range, rather it tells gretl to
_interpret_ your data as starting at a particular "historical
time" observation. I take it that the algorithm you have in mind
is the following:
(1) Set the sample range to start at time t and end at time t + k.
Estimate a model over this range.
(2) Generate (and record in a useful form) a forecast for period
t + k + 1.
(3) Have we reached the end of the available data? If so, stop;
if not, let t <- t + 1 and go to step (1).
If my interpretation is correct, the way to do this in gretl is
something like the following. (For the sake of concreteness I'm
referencing a particular dataset and model).
<script>
open data9-7
smpl ; 1988:4 # reserve 2 years of quarterly observations
series fc = NA # create a blank series to hold the forecasts
series fci
loop i=1989:1..1990:4
ols 1 0 2 3 4
fcast $i $i fci # generate period-specific forecast
fc[$i] = fci[$i] # and write it into the 'holder'
smpl +1 +1 # advance the start and end of the
# sample by 1 observation
end loop
print 1 fc --byobs # print the results
</script>
Note that you can use the --quiet option with the ols command if
you don't want to see the output from all the models.
Allin Cottrell
5959
days inactive
5974
days old
3 comments
3 participants
participants (3)
-
Allin Cottrell -
artur -
Rebecca Zhang