6:37 p.m.
On Fri, 29 Apr 2011, Ignacio Diaz-Emparanza wrote:
El 28/04/11 19:36, Ferenci Tamas escribió:
> Hello all,
>
> First of all: thank you Allin! It works perfectly, exactly that I was
> thinking of. I can even edit the plot gretl-stlye after running the
> script, very nice. (And I'm also a bit relieved that not I was missing
> a GUI feature for this...) By the way, I really appreciate the way you
> help everyone on this list, thank you again!
>
> Just out of curiosity: what is the rationale behind limiting the GUI
> to binary variables? Enabling the usage of any discrete variable
> would in no way alter the possibility of still using a binary variable
> (it would be a fully "backward compatible" change, so to say), and it
> also doesn't seem to be too complicated...
I am using Allin's script as well, for a cross-section model and a
discrete variable with three categories. I also vote for extending the
"XY with factor separation" plot to allow for more categories, of course
when our dear developers have time and are bored (no high priority).
OK, now in CVS and snapshots you can use any discrete variable,
not just a 0/1 dummy, with gnuplot and the --dummy option.
In the same line of business I've added a --factorized option for
the boxplot command, and enhanced the GUI boxplot options (e.g.,
on right-clicking a variable and selecting Boxplot).
Allin
9:02 p.m.
New subject: scatter plot with non-binary factor separation
On Fri, 29 Apr 2011, Allin Cottrell wrote:
In the same line of business I've added a --factorized option
for
the boxplot command, and enhanced the GUI boxplot options (e.g.,
on right-clicking a variable and selecting Boxplot).
The other day I spent more or less 3 hours trying to implement this very
same thing (and gave up eventually). Not only you're an incredible coder,
you've also become a mind-reader. Scary.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
10:01 a.m.
New subject: scatter plot with non-binary factor separation
El 29/04/11 20:37, Allin Cottrell escribió:
OK, now in CVS and snapshots you can use any discrete variable,
not just a 0/1 dummy, with gnuplot and the --dummy option.
In the same line of business I've added a --factorized option for
the boxplot command, and enhanced the GUI boxplot options (e.g.,
on right-clicking a variable and selecting Boxplot).
Allin
I have tried the new factorized plot, an I am obtaining this error:
gnuplot PIB08 pob08 Cod_prov --dummy
set xrange [1.797693135e+308:1.797693135e+308]
^
"/home/ignacio/.gretl/gpttmp.XpUlk9", line 17: undefined value
Cod_prov is a variable with three integer values, that I marked as
discrete in the "attributes" dialog. I must say I am obtaining this
error when I work with a dataset formed by 431 variables. I prepared a
dataset with only this three variables for sending to this list, but I
surprisingly found that the command works OK for this small dataset.
PIB08 and pob08 are variables with rather "normal" values and without
any missing.
By other side, I had no problem with the new "factorized boxplot"
although using the large dataset.
--
Ignacio Diaz-Emparanza
DEPARTAMENTO DE ECONOMÍA APLICADA III (ECONOMETRÍA Y ESTADÍSTICA)
UPV/EHU Avda. Lehendakari Aguirre, 83 | 48015 BILBAO
T.: +34 946013732 | F.: +34 946013754
www.ea3.ehu.es
5:45 p.m.
New subject: scatter plot with non-binary factor separation
Hello all,
I download the snapshot, and it works! Integrated into the GUI,
fantastic. (I mean: I can't think of any other software where a
feature idea would be implemented so rapidly.) Thank you again!
A short side note. I believe that such factorization is most useful
(in addition to boxplot, you've just implemented!) in case of
histograms and kernel density estimations, where the respective
diagrams beneath (in case of histograms) or superimposed on each other
(kernel density estimates) are very effective to visually assess
whether there is significant difference in the expected value (and
perhaps variance) of a variable between the groups. (Kind of a
graphical way to the same problem which is analytically addressed by
the two sample t-test or ANOVA.) I called it a "side note" because
this most often arises in biostatistics, psychometrics etc. as I see
(that is, more experimental applications of statistics) and not in
econometrics - so I completely understand if you decide intentionally
not to implement this in gretl, irrespectively of your free time.
Thank you again for you attention,
Tamás
2011. április 29., 20:37:59, írtad:
On Fri, 29 Apr 2011, Ignacio Diaz-Emparanza wrote:
> El 28/04/11 19:36, Ferenci Tamas escribió:
> > Hello all,
> >
> > First of all: thank you Allin! It works perfectly, exactly that I was
> > thinking of. I can even edit the plot gretl-stlye after running the
> > script, very nice. (And I'm also a bit relieved that not I was missing
> > a GUI feature for this...) By the way, I really appreciate the way you
> > help everyone on this list, thank you again!
> >
> > Just out of curiosity: what is the rationale behind limiting the GUI
> > to binary variables? Enabling the usage of any discrete variable
> > would in no way alter the possibility of still using a binary variable
> > (it would be a fully "backward compatible" change, so to say), and it
> > also doesn't seem to be too complicated...
>
> I am using Allin's script as well, for a cross-section model and a
> discrete variable with three categories. I also vote for extending the
> "XY with factor separation" plot to allow for more categories, of course
> when our dear developers have time and are bored (no high priority).
OK, now in CVS and snapshots you can use any discrete variable,
not just a 0/1 dummy, with gnuplot and the --dummy option.
In the same line of business I've added a --factorized option
for
the boxplot command, and enhanced the GUI boxplot options (e.g.,
on right-clicking a variable and selecting Boxplot).
Allin
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
4994
days inactive
5003
days old
3 comments
4 participants
participants (4)
-
Allin Cottrell -
Ferenci Tamas -
Ignacio Diaz-Emparanza -
Riccardo (Jack) Lucchetti