R help - Nov 2000 - FW: boxplot grouped by two variables: general issue

> On Tue, 21 Nov 2000, Vadik Kutsyy wrote:
>
> > Is there a quick way to make boxplots groups by two
> variables?  By that
> > I mean, that if x axes have values
("A","B","C"), than at
> each value
> > there would be a few boxplots each for a value of second
> variable (say
> > ("1","2","3")).
>
> If I understand you correctly you want to try the
> "interaction" function
> in a boxplot.  Your data needs to be organised into a data.frame with
> appropriate factors.  Here's a simple example ...
I'd rather prefer to have something like

  coplot(breaks | wool * tension, data = warpbreaks, panel=function(x, y,
...)boxplot(x))

but there are three problems:

1) coplot doesn't seem to support univariate panel functions
2) coplot doesn't seem to support synchronization options for xlim= and
ylim= across panels
3) plot functions appear to be EITHER panel functions OR high level plot
functions. Needing two versions for each plot function is a little bit
complicated. Rather have ONLY panel functions and a one-panel high level
wrapper.

Here are my questions:

a) Any plans to implement trellis?
b) Are there any other solutions in R which offer simple trellis
functionality?
c) Do you think, attacking these problems leads to reinventing trellis with
all it's complexity?

If the answers turn out to be c(a=NO,b=NO,c=NO) I would like  to ask in more
detail:

Is it possible to write plot functions to properly behave as high level AND
as panel functions?
Are there any issues that NEED to be addressed beyond 1) to 3) ?
Any recommendations? Any related work in progress? Any volunteers?

If nobody else is attacking this, I will think about a solution (r-devel:
AFTER the strwidth() issue has been resolved)


Jens


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

On Thu, Nov 23, 2000 at 12:07:55PM +0100, Jens Oehlschl?gel
wrote:
> 
> Here are my questions:
> 
> a) Any plans to implement trellis?
	Yes, in fact work is in progress.  Paul Murrell is working on
	generalized layout facilities which will provide the necessary
	support for multipanel conditioning (and more).  It is probably
	premature to talk about a time frame for this, but can report
	that I have seen trellis-like plots on Paul's screen.

	Of course, layout support is just the low-level part of trellis,
	but I think the higher level parts could come comparatively
	quickly once the layout support is in place.  (Issues are things
	like shingles and banking).
> b) Are there any other solutions in R which offer simple trellis
> functionality?
	Deepayan Sarkar has been working on modifications to coplot
	to make it more trellis-like.  This could well be placed over
	Paul's layout code at some point.
> c) Do you think, attacking these problems leads to reinventing trellis with
> all it's complexity?
> 
> If the answers turn out to be c(a=NO,b=NO,c=NO) I would like  to ask in
more
> detail:
> 
> Is it possible to write plot functions to properly behave as high level AND
> as panel functions?
	I want to look issues like this.  Some kind of partitioning of
	functionality is clearly needed.  It should be possible to
	access plots at a variety of levels

	 (i)	High level,
	 (ii)	Purely computational (no plotting).
		Some methods accept a plot=FALSE argument.

	In addition, it seems like we need:

	 (iii)	an intermediate level which allows the embedding
		of plots within other plots.

	Examples of this last are trellis panels and panels of plots
	superimposed on maps.
> Are there any issues that NEED to be addressed beyond 1) to 3) ?
> Any recommendations? Any related work in progress? Any volunteers?
	A useful contribution would be a look at how to package the
	functionality of the higher levels of trellis.	There are general
	issues of compatibility and extension, as well as specific
	functionality that needs to be provided.  The "for the record"
	sections of Cleveland's latest book are a good source for
	the latter.

Ross
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._