[R] Lattice plot questions

[Deepayan Sarkar]

Let's consider this simulated data:

foo <- data.frame(resp = do.call("c", lapply(as.list(rep(6, 12)),
                  function(x) sort(rnorm(x)))),
                  week = rep(2*0:5, 12),
                  id = factor(rep(1:12, each = 6)))

Does the following give you what you want ?

xyplot(resp ~ week | id, foo,

       panel = function(x, y, col = sup.sym$col, panel.number, ...) {

           ## this part accesses the data frame directly, and needs to
           ## be adjusted accordingly if the name of the variables
           ## therein change. This draws the part meant for the whole
           ## data set

           panel.superpose(x = foo$week, y = foo$resp,
                           subscripts = TRUE, groups = foo$id,
                           type = 'l', col = "black")

           ## this part draws the colored lines, each overwriting one
           ## of the background lines

           sup.sym <- trellis.par.get("superpose.symbol")
           col.this <- col[1 + ((panel.number - 1) %% length(col))]
           panel.xyplot(x, y, type = 'b', col = col.this, 
                        lwd = 2, cex = 1.2, ...)
           
       })

This will by default use the colors in superpose.symbol, but you can override 
that by specifying your own color as an argument to xyplot, perhaps something 
like 

xyplot(resp ~ week | id, foo, col = c("red", "yellow", "blue"),
       panel = ....

I personally don't see the point of using different colors though (that would 
seem useful to me only if the different colors appeared within the same 
panel).

There's no restriction on the length of the elements of 
trellis.par.get("superpose.symbol"). 7 (which is just as good a number as any 
other) is the default choice, inspired by Trellis in S. Set it to whatever 
you want with trellis.par.set() or lset() (or just supply the colors to 
xyplot(), which is usually more convenient for specific tasks like this).

Names like "navy blue" are just a convenience, and I doubt that all 256^3 
colors in the RGB space have been named, so trying to get a name for 
"#ff7f00" may not get you anywhere. R has a function col2rgb() that converts 
colors to RGB, e.g.

> col2rgb("#ff7f00")
      [,1]
red    255
green  127
blue     0
> col2rgb("navyblue")
      [,1]
red      0
green    0
blue   128

If you are more comfortable with named colors, use them--- R accepts them 
everywhere. A list of names it recognizes is produced by 

> colors()

The help pages for colors and col2rgb have more information.

HTH,

Deepayan

On Friday 29 August 2003 11:14, Paul, David A wrote:
> Win2k, R1.7.1:
>
> I am currently working with some growth curve data from a
> biotoxicology experiment.  Each of 12 subjects had their blood
> drawn at 0, 2, 4, 6, 8, and 10 weeks.  For the purposes of the
> project, it would be helpful if I were able to do the following:
>
> 	a.  Produce 12 panels, each displaying the *same* data, with
> 		the "strip" at the top of a particular panel showing
> 		exactly one of the subject id's
> 	b.  For a particular panel, plot the time course data for
> 		the subject whose id appears in the strip in some
> 		really obvious color (like yellow, red, or blue),
> 		with circles at each time point, and connecting the
> 		"dots".  In the same panel, those subjects whose id's
> 		did NOT appear in the strip would ideally have their
> 		data plotted in black, possibly with no symbols at
> 		any of the time points, but still connecting the
> 		"dots".
>
> I apologize if the description of what is needed is vague.
> It is possible to address (a) by duplicating the data set
> twelve times and cbind()ing a factor variable so that each
> of the duplicated data sets is associated with exactly one
> of the subject id's.  I have done this, and the code
>
> subjects <- foo$subject.id
> superpose.symbol <- trellis.par.get("superpose.symbol")
> xyplot(log.response ~ week | factor.variable, data = foo,
> 	panel = panel.superpose,
> 	groups = subjects,
> 	key = list(space = "top", columns = 6, transparent = TRUE,
> 			text = list(levels(subjects)),
> 			points = Rows(superpose.symbol,1:7))
>        )
>
> works, with "factor.variable" corresponding to the cbind()ed
> factor variable.  Unfortunately, I have been unable to figure
> out how to do (b).  I would also like to determine a more
> memory efficient way of doing (a) -- even though my current
> data set is small (only 10 weeks of data per subject), that will
> change.
>
> Also, an examination of "superpose.symbol" reveals the
>
> following:
> > superpose.symbol <- trellis.par.get("superpose.symbol")
> > superpose.symbol
>
> $cex
> [1] 0.8 0.8 0.8 0.8 0.8 0.8 0.8
>
> $col
> [1] "#00ffff" "#ff00ff" "#00ff00" "#ff7f00" "#007eff" "#ffff00" "#ff0000"
>
> $font
> [1] 1 1 1 1 1 1 1
>
> $pch
> [1] "o" "o" "o" "o" "o" "o" "o"
>
>
> This seems to indicate that lattice graphics are restricted to
> no more than seven colors, though I cannot believe that this is true.
> Where might I search/read/learn about the available colors for
> lattice plots and how to denote them?  (I find "#ff7f00" to be
> a little cryptic, for example.)  Is there a table that relates
> these ?hexadecimal? numbers to descriptions like "navy blue"
> for Win2k?
>
> Finally, is it possible to cause superpose.symbol to default to
> more than 7 $cex/$col/$font/$pch values?  Since I am working with
> 12 subjects, plots that assign a different color to each wind up
> duplicating colors and symbols at some point.
>
>
> Much thanks in advance,
>   david paul
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help