Assorted little bugs
by Sven Schreiber
Hi,
during a bug-hunting session which will lead to an update of py4gretl
shortly I came across the following quirks and issues in gretl (with Win
snapshot Apr 17):
1) x = (obs>1991:1) isn't working ("expected ')' but found ':'" or
something like that -- this is with quarterly data).
2) In a "new variable" dialog, x=0 produces scalar, which is correct,
but having to type series x = 0 in the GUI is suboptimal IMHO. Maybe the
desired data type could be set by radio buttons in the dialog?
3) in "edit values" dialog, the view doesn't follow cursor movements
(limitation of GTK?)
4) old bug resurfacing (?) : selected variable in window isn't used if
edit variable attributes is chosen from menu
5) in edit matrix dialog (opened from session icons window), the name of
the edited matrix isn't displayed (but should be)
6) are named lists of variables visible somewhere, like matrices in the
session window?
7) Is there any way to start gretl with a console window showing error
messages on Windows? (Even if I start gretlw32.exe from a console, gretl
starts but the prompt returns "promptly" --pun intended--, and so I
don't get any error messages like when I type "gretl" in a Linux
console. In my case, I would like to see the the error messages from
python that was started by gretl/py4gretl.)
8) It seems there is a problem if named lists start with a number;
example from a console session:
? list 4dim
4dim: GesamtDEUUNR deinfl dewreal deprod
? list fourdim
fourdim: GesamtDEUUNR deinfl dewreal deprod
? matrix test2 = py4gretl_vecm(1, fourdim, 2, 1, 3, 0, null, null)
Befehl 'p4g_call' unbekannt
# (that's ok, this error is expected)
? matrix test2 = py4gretl_vecm(1, 4dim, 2, 1, 3, 0, null, null)
> matrix test2 = py4gretl_vecm(1, 4dim,
Das Symbol 'dim' ist nicht definiert
Unbekannter Variablenname im Befehl
# (but this is bogus, should be the same error as before)
That's all ;-)
cheers,
sven
17 years, 2 months
Johansen's test
by Mariusz Doszyń
Hello,
I've got a few questions about results of Johansen's test in Gretl. I don't know how exactly these results are computed. I've calculated eigenvalues according to the popular formula proposed by Johansen: det[lambda*I-[(S11)-1]*(S10)*[(S00)-1]*(S01)] - and results are different (0,58 and 0,23; in Gretl: 0,67 and 0,17). Do you know why? Could you also tell me how two first matrices in output were computed. They are called: beta (cointegrating vector) and alpha (fitted vectors)? I'm familiar with the idea which is behind, but I'd like to be able to compute them myself... And the last question. Do you know any sources on the internet where VAR, VECM models and connected methods are (clearly) described. I'm sorry for bothering you with such a problems but I've got some problems with access to literature now. Thanks a lot for your help.
Best wishes,
Mariusz Doszyń
Poland
17 years, 2 months
Re:Re: [Gretl-users] following Sven's hunting
by ab.news
Sven Schreiber a écrit :
> ab.news schrieb:
>
>> Suggestion
>> For those of us who have to deal with a bi(multi)lingual
>> teaching/terms would it be possible to build in a language
>> switch menu to avoid shuttling Dos-Gretl sessions?
>>
>>
>
> I'm not a gretl developer, but that sounds difficult to
implement.
> (Although I agree it would be nice.) Do you know any other
program that
> can do that?
>
> -sven
> _______________________________________________
> Gretl-users mailing list
> Gretl-users(a)lists.wfu.edu
> http://lists.wfu.edu/mailman/listinfo/gretl-users
>
I don't know any econometric program doing this. But Gretl is
unique isn't it?
However, there is plenty of other softwares doing it (eg.
Advanced Grapher --a stuff that plots functions' graphs--,
just to mention one).
The idea behind is very simple and the implementation doesn't
need a time-expensive effort I guess (at least in Visual Basic
programing : I tried it myself).
The language file would be a plain ASCII text file provided
with the program. The advantage is that anyone can easily
modify e ".txt" file and obtain his own "language file"
without waiting an "official" transalation or even to correct
or tune it.
Actually, doing this with Gretl is somehow troublesome: I
guess you have to modify the ".mo" file and I'm not sure if
one have to re-compile Gretl for changes taking effect!
Anyway, I have no idea how one can make Gretl speak his language.
artur
Créez votre adresse électronique prenom.nom(a)laposte.net
1 Go d'espace de stockage, anti-spam et anti-virus intégrés.
17 years, 3 months
following Sven's hunting
by ab.news
following Sven's hunting (on Windows snapshot April 22):
1. the command "genr dummy = (t=10)" doesn't produce a dummy
variable as indicated in the User's guide (july 2006 version)
2. while executing the following script :
nulldata 4100
setobs 1 1 --time-series
genr X = normal()+2
gnuplot x --time-series
double-click on "X" to display X values and you'll see that
the "time series graph" icon has disappeared from the toolbar.
It does not happen with less than 4100 observations.
Suggestion
For those of us who have to deal with a bi(multi)lingual
teaching/terms would it be possible to build in a language
switch menu to avoid shuttling Dos-Gretl sessions?
thanks a lot
Créez votre adresse électronique prenom.nom(a)laposte.net
1 Go d'espace de stockage, anti-spam et anti-virus intégrés.
17 years, 3 months
Bug report for gretl's interaction with R
by andreas.karlsson＠ltv.se
Hi,
I am using gretl on Windows XP. I have found a bug in gretl regarding
gretl's interaction with R:
When I have installed both R and gretl, in Tools > Preferences >General >
Programs > Command to launch GNU R it says "RGUI.exe" in the box.
First bug: Using Tools > Start GNU R results in an error message: "File not
found."
Second bug: When clicking on Tools > Preferences >General > Programs >
Command to launch GNU R > Browse... and adding the whole path to RGUI.exe
in the box , the Tools > Start GNU R works perfectly and starts R. However,
when closing gretl and restarting it, the path has disappeared and it says
only "RGUI.exe" in the box again. And trying to start R again results in
the first bug again.
Best regards
Andreas Karlsson
17 years, 3 months
Ang. Re: [Gretl-users] Re: Re: Bug for XY scatterplot
by andreas.karlsson＠ltv.se
Thanks a lot!
Best regards
Andreas
cottrell(a)wfu.edu @ INTERNET skrev 2007-04-24 22:38:21 :
> On Tue, 24 Apr 2007, andreas.karlsson(a)ltv.se wrote:
>
> > Could you please implement it for "Fitted line > inverse" too?
>
> OK, done in CVS.
>
> -- Allin Cottrell
> Department of Economics
> Wake Forest University, NC
> _______________________________________________
> Gretl-users mailing list
> Gretl-users(a)lists.wfu.edu
> http://lists.wfu.edu/mailman/listinfo/gretl-users
17 years, 3 months
Re: Re: Bug for XY scatterplot
by andreas.karlsson＠ltv.se
Hi,
Could you please implement it for "Fitted line > inverse" too?
Thanks in advance.
Best regards
Andreas Karlsson
>> A bug for XY scatterplots in gretl:
>>
>> 1. Plot a XY scatterplot from View > Graph specified vars > X-Y
scatter...
>>
>> 2. Click on the graph to obtain the pop-up menu and choose Edit > Main >
>> Fitted line > quadratic > OK
>>
>> 3. Click on the graph to obtain the pop-up menu: The "OLS estimates"
menu
>> entry has disappeared.
>
>Well, I wouldn't exactly say it's a bug; rather a non-implemented
>feature. However, that feature is now implemented in CVS.
>
>Allin.
17 years, 3 months
correlation matrix
by ab.news
Hi everybody!
In the correlation matrix window, in the presence of missing
values, does the 5% critical value make any sense since the
number of observations "n" is not the same for all pairwise
correlations? I guess -- when some some values are missing--,
we'll need either a 5% critical value for each correlation
coefficient or a matrix correlation computed on a listwise
basis, meaning that the same observations are used throughout
the dataset.
Artur
Créez votre adresse électronique prenom.nom(a)laposte.net
1 Go d'espace de stockage, anti-spam et anti-virus intégrés.
17 years, 3 months
zero values
by Marco Marini
Hi,
I created an harmonic with the following command:
a1=sin(pi*time)
where time is a trend series. I expected a zero vector,
but the program returns something which is very close to
zero but not exactly, as shown in the tabulation
Obs b2
1981:1 1.22461e-016
1981:2 -2.44921e-016
1981:3 3.67382e-016
...
and in the graph, which is not a straight line at y=0.
I understand this is due to the numerical accuracy at which the program
works,
but I wonder if it's possible to set an option to show 0 whenever
values are lower than a fixed level (say 10^-5).
Thanks,
Marco Marini
Istat
17 years, 3 months
John Finlayson/NSWAG is out of the office.
by john.finlayson＠dpi.nsw.gov.au
I will be out of the office starting 20/04/2007 and will not return until
28/05/2007.
I can be contacted either by email finlaysonjd(a)yahoo.co.uk or by mobile
0405 641 181 otherwise I will reply when I return.
17 years, 3 months
