5:49 a.m.
Hello gretl users and programmers,
I would like to request a feature for plotting matrices -- of course only
if it would not be too difficult to implement and makes some sense to you
;-)
Until now one has to define each column which one wants to plot. Could it
be possible to make an option available like
----------------
gnuplot 1..5 ---with-lines --time-series --matrix=X
----------------
to make it easier to plot several columns in a sequence (1 to 5) at once?
I ask for this since i've got to generate some figures via a loop command
and the underlying matrices have different dimensions:
----------------
loop for i=1..models --quiet
TR_fw_$i <- gnuplot 1..k --with-lines --time-series --matrix=trace_fw_$i
endloop
----------------
Maybe somebody has another simple solution for this?
Best,
Artur
6:01 a.m.
On 12/05/2011 11:49 AM, artur tarassow wrote:
Hello gretl users and programmers,
I would like to request a feature for plotting matrices -- of course
only if it would not be too difficult to implement and makes some sense
to you ;-)
Until now one has to define each column which one wants to plot. Could
it be possible to make an option available like
----------------
gnuplot 1..5 ---with-lines --time-series --matrix=X
----------------
to make it easier to plot several columns in a sequence (1 to 5) at once?
I ask for this since i've got to generate some figures via a loop
command and the underlying matrices have different dimensions:
----------------
loop for i=1..models --quiet
TR_fw_$i <- gnuplot 1..k --with-lines --time-series --matrix=trace_fw_$i
endloop
----------------
Maybe somebody has another simple solution for this?
Write the column indices into a string and insert the string in your
gnuplot command. Something like (totally untested):
sprintf strColindices "%d", seq(1,cols(mymatrix))
gnuplot @strColindices --with-lines <etc>
hth,
sven
6:30 a.m.
Nice idea, sven. Thank you for your reply.
But unfortunately this does not work properly, since the string printed
looks like
-------
1234
-------
but gnuplot needs for seperating the columns a space between the numbers:
----------
1 2 3 4
----------
Best,
Artur
2011/12/5 Sven Schreiber <svetosch(a)gmx.net>
On 12/05/2011 11:49 AM, artur tarassow wrote:
> Hello gretl users and programmers,
>
> I would like to request a feature for plotting matrices -- of course
> only if it would not be too difficult to implement and makes some sense
> to you ;-)
> Until now one has to define each column which one wants to plot. Could
> it be possible to make an option available like
>
> ----------------
> gnuplot 1..5 ---with-lines --time-series --matrix=X
> ----------------
>
> to make it easier to plot several columns in a sequence (1 to 5) at once?
>
> I ask for this since i've got to generate some figures via a loop
> command and the underlying matrices have different dimensions:
>
> ----------------
> loop for i=1..models --quiet
> TR_fw_$i <- gnuplot 1..k --with-lines --time-series
--matrix=trace_fw_$i
> endloop
> ----------------
>
> Maybe somebody has another simple solution for this?
>
Write the column indices into a string and insert the string in your
gnuplot command. Something like (totally untested):
sprintf strColindices "%d", seq(1,cols(mymatrix))
gnuplot @strColindices --with-lines <etc>
hth,
sven
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
6:32 a.m.
On Mon, 5 Dec 2011, Sven Schreiber wrote:
Write the column indices into a string and insert the string in your
gnuplot command. Something like (totally untested):
sprintf strColindices "%d", seq(1,cols(mymatrix))
gnuplot @strColindices --with-lines <etc>
In other words:
<hansl>
T = 20
matrix x_1 = mnormal(T, 3)
matrix x_2 = mnormal(T, 2)
matrix x_3 = mnormal(T, 6)
loop i=1..3 --quiet
colsel = ""
c = cols(x_$i)
loop j=1..c --quiet
sprintf colsel "%s $j", colsel
end loop
gnuplot @colsel --matrix=x_$i --output=display
end loop
</hansl>
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
6:45 a.m.
On Mon, 5 Dec 2011, Sven Schreiber wrote:
sprintf strColindices "%d", seq(1,cols(mymatrix))
gnuplot @strColindices --with-lines <etc>
A technical note to Sven's idea: the problem is not the space (or lack
thereof): a print format like "%3d" would have solved the problem; the
problem is that when you "printf" a matrix, a newline is automtically
appended, which confuses the gnuplot command.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
9:05 a.m.
Thank you both Sven and Riccardo for your help.
I did now in this way, and it works fine.
----------------------------------------------
loop i=1..models --quiet
colind = ""
k = nelem(mod$i)
loop j=1..k #--quiet
sprintf colind "%s $j", colind
end loop
TR_fw_$i <- gnuplot @colind --with-lines --time-series
--matrix=trace_fw_$i
end loop
----------------------------------------------
Best,
Artur
2011/12/5 Riccardo (Jack) Lucchetti <r.lucchetti(a)univpm.it>
On Mon, 5 Dec 2011, Sven Schreiber wrote:
sprintf strColindices "%d", seq(1,cols(mymatrix))
> gnuplot @strColindices --with-lines <etc>
>
A technical note to Sven's idea: the problem is not the space (or lack
thereof): a print format like "%3d" would have solved the problem; the
problem is that when you "printf" a matrix, a newline is automtically
appended, which confuses the gnuplot command.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/**lucchetti<http://www.econ.univpm.it/lucche...
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
10:34 a.m.
On Mon, 5 Dec 2011, artur tarassow wrote:
Thank you both Sven and Riccardo for your help.
I did now in this way, and it works fine.
----------------------------------------------
loop i=1..models --quiet
colind = ""
k = nelem(mod$i)
loop j=1..k #--quiet
sprintf colind "%s $j", colind
end loop
TR_fw_$i <- gnuplot @colind --with-lines --time-series
--matrix=trace_fw_$i
end loop
----------------------------------------------
This may not be directly relevant to the task you describe,
but it's not too far distant: in CVS and snapshots, the
"yvars" and "xvar" arguments to the gnuplot command can now be
left implicit when a matrix is specified via the --matrix
option.
If --time-series is given the default is to plot all columns
of the matrix against time, otherwise the last column of the
matrix is taken as "xvar" and the others as "yvars".
Of course, you can still specify the columns to plot if you
wish.
Allin Cottrell
11:31 a.m.
On Mon, 5 Dec 2011, Allin Cottrell wrote:
This may not be directly relevant to the task you describe, but
it's not too
far distant: in CVS and snapshots, the "yvars" and "xvar" arguments
to the
gnuplot command can now be left implicit when a matrix is specified via the
--matrix option.
for example, if you want to see an nice bunch of random walks:
<hansl>
X = cum(mnormal(40,3)) ~ seq(1,40)'
gnuplot --matrix=X --output=display --with-lines
</hansl>
If --time-series is given the default is to plot all columns of the matrix
against time, otherwise the last column of the matrix is taken as "xvar" and
the others as "yvars".
or
<hansl>
X = cum(mnormal(40,3))
gnuplot --matrix=X --output=display --with-lines --time-series
</hansl>
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
2:09 p.m.
2011/12/5 Riccardo (Jack) Lucchetti <r.lucchetti(a)univpm.it>
On Mon, 5 Dec 2011, Allin Cottrell wrote:
This may not be directly relevant to the task you describe, but it's not
> too far distant: in CVS and snapshots, the "yvars" and "xvar"
arguments to
> the gnuplot command can now be left implicit when a matrix is specified via
> the --matrix option.
>
for example, if you want to see an nice bunch of random walks:
<hansl>
X = cum(mnormal(40,3)) ~ seq(1,40)'
gnuplot --matrix=X --output=display --with-lines
</hansl>
> If --time-series is given the default is to plot all columns of the
> matrix against time, otherwise the last column of the matrix is taken as
> "xvar" and the others as "yvars".
>
or
<hansl>
X = cum(mnormal(40,3))
gnuplot --matrix=X --output=display --with-lines --time-series
</hansl>
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/**lucchetti<http://www.econ.univpm.it/lucche...
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
That's good know!
Btw, is there a reason why the "scatters" command does not work for
matrices?
Best,
Artur
2:51 p.m.
2011/12/5 Allin Cottrell <cottrell(a)wfu.edu>
On Mon, 5 Dec 2011, artur tarassow wrote:
> Btw, is there a reason why the "scatters" command does not work for
> matrices?
Yes: nobody has hooked it up to do so ;-)
It wouldn't be terribly difficult, but it wouldn't be trivial
and nobody has requested it.
Allin Cottrell
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
Ok let me try to convince you ;-)
A more concrete example is the plotting of the error-correction terms
(vecm, ec plot). It would be nice to have a diagram for each cointegration
relation. Since these are stored a matrix (as I assume), this would (maybe)
require some matrix-plotting-options for the "scatters" command :-)
2:12 p.m.
On Mon, 5 Dec 2011, artur tarassow wrote:
2011/12/5 Allin Cottrell <cottrell(a)wfu.edu>
> On Mon, 5 Dec 2011, artur tarassow wrote:
>
>> Btw, is there a reason why the "scatters" command does not work for
>> matrices?
>
> Yes: nobody has hooked it up to do so ;-)
>
> It wouldn't be terribly difficult, but it wouldn't be trivial
> and nobody has requested it.
Ok let me try to convince you ;-)
A more concrete example is the plotting of the
error-correction terms (vecm, ec plot). It would be nice to
have a diagram for each cointegration relation. Since these
are stored a matrix (as I assume), this would (maybe)
require some matrix-plotting-options for the "scatters"
command :-)
OK, since enhancement of graphing has been a sub-plot of
working towards gretl 1.9.7, I've added a --matrix=whatever
option to the "scatters" command. You can give the x and y
specifications with series, but if you omit them you get a
plot of all columns of the matrix against time (or an index
variable). While I was at it I raised the max number of plots
in "scatters" to 16. Example:
<hansl>
matrix = mnormal(40,16)
scatters --with-lines --output=display --matrix=m
</hansl>
Allin Cottrell
2:57 p.m.
OK, since enhancement of graphing has been a sub-plot of
working towards gretl 1.9.7, I've added [...]
Hi Allin, is there any chance to
have my former wish accomplished ? :)
On Sun, 16 Oct 2011, artur bala wrote:
Is there a way to plot in the same graph pairwise series (xi, yi)?
I'm thinking to something like this:
list X = x1 x2
list Y = y1 y2
gnuplot Y X # plotting (x1,y1) and (x2,y2)
gretl's interface to gnuplot
supports only a single x axis,
and a single x variable. But you can write a graph
specification for gnuplot that uses 2 x variables, and 2 x
axes if you wish.
Dump your data into a plain text file, e.g.
# y1 x1 y2 x2
2 1 1000 100
4 2 800 200
6 3 600 300
8 4 400 400
10 5 200 500
and do the plot something like this:
<gnuplot>
set ytics nomirror
set y2tics
set xtics nomirror
set x2tics
plot 'twindata' using 1:2 axes x1y1 , \
'twindata' using 3:4 axes x2y2
</gnuplot>
Note that you have a gnuplot editor available in gretl under
/File/Script files/New script/gnuplot script.
7:12 p.m.
On Tue, 6 Dec 2011, artur bala wrote:
> OK, since enhancement of graphing has been a sub-plot of
> working towards gretl 1.9.7, I've added [...]
Hi Allin, is there any chance to have my former wish accomplished ? :)
That is (Artur's wish):
> Is there a way to plot in the same graph pairwise series (xi,
yi)?
> I'm thinking to something like this:
> list X = x1 x2
> list Y = y1 y2
> gnuplot Y X # plotting (x1,y1) and (x2,y2)
Hmm. The first point is that "gnuplot Y X" wouldn't work: "Y X"
would just be one big list from gretl's point of view; we'd need new
syntax such as "gnuplot Y ; X" to separate the x-axis and y-axis
lists.
But more generally, I'm skeptical. There an obvious case for
plotting two (or more) y-variables against one x, but combining two
plots with different y's and different x's in one frame? I think
this is likely to be more confusing that elucidating to the viewer.
Well, I suppose I can imagine cases where one would like to do this
(though it's not something I've ever wanted to do), but it's
sufficiently uncommon that I'm inclined to repeat my previous
advice: you can do it in a gnuplot command file (without too much
difficulty) if you really want to.
If you can make a good argument for making this a built-in option in
gretl I'll think about it some more, but right now it seems more
trouble than it's worth.
Allin
3:40 a.m.
On Tue, 6 Dec 2011, Allin Cottrell wrote:
Well, I suppose I can imagine cases where one would like to do this
(though it's not something I've ever wanted to do), but it's
sufficiently uncommon that I'm inclined to repeat my previous
advice: you can do it in a gnuplot command file (without too much
difficulty) if you really want to.
Let me add: not only you can do it via gnuplot syntax, you can also
produce the two plots separately and arrange them side by side (or any
other way you like) in your document: I've almost always found the
"tabular" environment more than adequate to the purpose.
More generally, I agree that one of gretl's strong points (which I can
take absolutely no credit for) is its very very nice way of interacting
with gnuplot, but I'm strongly against going over the top and acting as if
gretl was some kind of gnuplot frontend.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
5:23 a.m.
OK, since enhancement of graphing has been a sub-plot of
working towards gretl 1.9.7, I've added a --matrix=whatever
option to the "scatters" command. You can give the x and y
specifications with series, but if you omit them you get a
plot of all columns of the matrix against time (or an index
variable). While I was at it I raised the max number of plots
i>n
"scatters" to 16. Example:
<hansl>
matrix = mnormal(40,16)
scatters --with-lines --output=display --matrix=m
</hansl>
Thank you for implementing this, Allin. The output looks very nice, and it
is good to have the possibility to plot more than 6 figures within scatters
(e.g. for drawing impulse-response-functions with some more variables
within one window).
But there seems to be a bug if one wants to save the output to *.eps. In
your example only the 13th sub-figure is saved.
Best,
Artur
2011/12/7 Riccardo (Jack) Lucchetti <r.lucchetti(a)univpm.it>
On Tue, 6 Dec 2011, Allin Cottrell wrote:
Well, I suppose I can imagine cases where one would like to do this
> (though it's not something I've ever wanted to do), but it's
> sufficiently uncommon that I'm inclined to repeat my previous
> advice: you can do it in a gnuplot command file (without too much
> difficulty) if you really want to.
>
Let me add: not only you can do it via gnuplot syntax, you can also
produce the two plots separately and arrange them side by side (or any
other way you like) in your document: I've almost always found the
"tabular" environment more than adequate to the purpose.
More generally, I agree that one of gretl's strong points (which I can
take absolutely no credit for) is its very very nice way of interacting
with gnuplot, but I'm strongly against going over the top and acting as if
gretl was some kind of gnuplot frontend.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/**lucchetti<http://www.econ.univpm.it/lucche...
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
10 a.m.
On Wed, 7 Dec 2011, artur tarassow wrote:
> OK, since enhancement of graphing has been a sub-plot of
> working towards gretl 1.9.7, I've added a --matrix=whatever
> option to the "scatters" command. You can give the x and y
> specifications with series, but if you omit them you get a
> plot of all columns of the matrix against time (or an index
> variable). While I was at it I raised the max number of plots
i>n "scatters" to 16. Example:
<hansl>
> matrix = mnormal(40,16)
> scatters --with-lines --output=display --matrix=m
</hansl>
Thank you for implementing this, Allin. The output looks very nice, and it
is good to have the possibility to plot more than 6 figures within scatters
(e.g. for drawing impulse-response-functions with some more variables
within one window).
But there seems to be a bug if one wants to save the output to *.eps. In
your example only the 13th sub-figure is saved.
Hmm, I see what you mean. I'll look into that.
Allin
3:53 p.m.
On Wed, 7 Dec 2011, Allin Cottrell wrote:
On Wed, 7 Dec 2011, artur tarassow wrote:
> [...]
> <hansl>
>> matrix = mnormal(40,16)
>> scatters --with-lines --output=display --matrix=m
> </hansl>
> [...]
>
> But there seems to be a bug if one wants to save the output to *.eps. In
> your example only the 13th sub-figure is saved.
Hmm, I see what you mean. I'll look into that.
I think this should be fixed now.
Allin
12:47 p.m.
Thank you for the fix. It works fine now! :)
Best,
Artur
2011/12/7 Allin Cottrell <cottrell(a)wfu.edu>
On Wed, 7 Dec 2011, Allin Cottrell wrote:
> On Wed, 7 Dec 2011, artur tarassow wrote:
>> [...]
>> <hansl>
>>> matrix = mnormal(40,16)
>>> scatters --with-lines --output=display --matrix=m
>> </hansl>
>> [...]
>>
>> But there seems to be a bug if one wants to save the output to *.eps. In
>> your example only the 13th sub-figure is saved.
>
> Hmm, I see what you mean. I'll look into that.
I think this should be fixed now.
Allin
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
5:35 a.m.
> Well, I suppose I can imagine cases where one would like to do
this
> (though it's not something I've ever wanted to do), but it's
> sufficiently uncommon that I'm inclined to repeat my previous
> advice: you can do it in a gnuplot command file (without too much
> difficulty) if you really want to.
Let me add: not only you can do it via gnuplot syntax, you can also
produce the two plots separately and arrange them side by side (or any
other way you like) in your document: I've almost always found the
"tabular" environment more than adequate to the purpose.
More generally, I agree that one of gretl's strong points (which I can
take absolutely no credit for) is its very very nice way of
interacting with gnuplot, but I'm strongly against going over the top
and acting as if gretl was some kind of gnuplot frontend.
Well, my initial purpose was to "translate" a couple of stata packages
(the svy- family) into gretl functions and by the way, wanted to add
some features the stata packages lack; I came out with this stuff of
plotting in the same graph many cumulative density functions of income -
commonly called the Pen's Parade - with population breakdown. I
absolutely agree that's an uncommon feature but that was just a wish to
provide users with a self-contained function producing different outputs
and graphs in a run.
By the way, the recent change in the gnuplot --matrix=whatever syntax
greatly simplified my script :)
Thank yoy again for your comments,
cheers,
artur
1:19 p.m.
On 12/05/2011 12:45 PM, Riccardo (Jack) Lucchetti wrote:
On Mon, 5 Dec 2011, Sven Schreiber wrote:
> sprintf strColindices "%d", seq(1,cols(mymatrix))
> gnuplot @strColindices --with-lines <etc>
A technical note to Sven's idea: the problem is not the space (or lack
thereof): a print format like "%3d" would have solved the problem; the
problem is that when you "printf" a matrix, a newline is automtically
appended, which confuses the gnuplot command.
Hm, I see. I seem to remember a special backspace character \b -- would
it work to add that to the string, in the hope of deleting the appended
newline character \n?
Alternatively, would it work to process strColindices further with:
strsub(strColindices,"\n","")
?
(BTW: I don't understand the help text for strsplit().)
Thanks,
Sven
1:37 p.m.
On Mon, 5 Dec 2011, Sven Schreiber wrote:
On 12/05/2011 12:45 PM, Riccardo (Jack) Lucchetti wrote:
> On Mon, 5 Dec 2011, Sven Schreiber wrote:
>
>> sprintf strColindices "%d", seq(1,cols(mymatrix))
>> gnuplot @strColindices --with-lines <etc>
>
> A technical note to Sven's idea: the problem is not the space (or lack
> thereof): a print format like "%3d" would have solved the problem; the
> problem is that when you "printf" a matrix, a newline is automtically
> appended, which confuses the gnuplot command.
>
Hm, I see. I seem to remember a special backspace character \b -- would
it work to add that to the string, in the hope of deleting the appended
newline character \n?
I don't think we support \b.
Alternatively, would it work to process strColindices further with:
strsub(strColindices,"\n","")
Seems like it probably should, but in fact the \n escape is
not recognized as a single character to be replaced.
(BTW: I don't understand the help text for strsplit().)
<hansl>
string s = "Three blind mice"
printf "The Trinity is %s in One.\n", strsplit(s, 1)
printf "I dig %s Lemon Jefferson.\n", strsplit(s, 2)
printf "The cat chased the %s.\n", strsplit(s, 3)
</hansl>
Allin
1:54 p.m.
On 12/05/2011 07:37 PM, Allin Cottrell wrote:
On Mon, 5 Dec 2011, Sven Schreiber wrote:
> On 12/05/2011 12:45 PM, Riccardo (Jack) Lucchetti wrote:
>> On Mon, 5 Dec 2011, Sven Schreiber wrote:
>>
>>> sprintf strColindices "%d", seq(1,cols(mymatrix))
>>> gnuplot @strColindices --with-lines <etc>
>>
>> A technical note to Sven's idea: the problem is not the space (or lack
>> thereof): a print format like "%3d" would have solved the problem; the
>> problem is that when you "printf" a matrix, a newline is automtically
>> appended, which confuses the gnuplot command.
>>
>
> Hm, I see. I seem to remember a special backspace character \b -- would
> it work to add that to the string, in the hope of deleting the appended
> newline character \n?
I don't think we support \b.
> Alternatively, would it work to process strColindices further with:
> strsub(strColindices,"\n","")
Seems like it probably should, but in fact the \n escape is
not recognized as a single character to be replaced.
I wonder then if it's really the best thing to append the \n when you
"printf" a matrix. I mean if it's really wanted it seems easy to add it,
but it looks like it's rather difficult to get rid of it...
But obviously this is not very important.
> (BTW: I don't understand the help text for strsplit().)
<hansl>
string s = "Three blind mice"
printf "The Trinity is %s in One.\n", strsplit(s, 1)
printf "I dig %s Lemon Jefferson.\n", strsplit(s, 2)
printf "The cat chased the %s.\n", strsplit(s, 3)
</hansl>
Thanks, maybe some example like this could be added to the help text.
cheers,
sven
2:49 p.m.
On Mon, 5 Dec 2011, Allin Cottrell wrote:
On Mon, 5 Dec 2011, Sven Schreiber wrote:
> Alternatively, would it work to process strColindices further with:
> strsub(strColindices,"\n","")
Seems like it probably should, but in fact the \n escape is not recognized as
a single character to be replaced.
Ah, that's because the string-substitution target, "\n", is
created on the fly as a string literal, in which escape
sequences are not recognized. This behavior makes it easier to
do things like
string filename = "c:\some-path\newfile"
But newlines can be substituted out of a string if you use
sprintf to create the target string, as in
<hansl>
sprintf str "some stuff\n"
printf "str = '%s'\n", str
sprintf target "\n"
string trimmed = strsub(str, target, "")
printf "trimmed = '%s'\n", trimmed
</hansl>
Allin
4657
days inactive
4660
days old
23 comments
5 participants
participants (5)
-
Allin Cottrell -
artur bala -
artur tarassow -
Riccardo (Jack) Lucchetti -
Sven Schreiber