[R] panel function in a conditioned lattice graphic

[Anthony Darrouzet-Nardi]

>On Saturday 29 May 2004 15:24, Anthony Darrouzet-Nardi wrote:
>  > I'm trying to use plotting character to encode the variable "block"
>  > from my dataset in a conditioned lattice graphic (R 1.9.0 on Mac OS
>  > 10.3.3). The data I'm using is the dataframe "dryoutcover" which is
>  > here (4k):
>  >
>  > http://anthony.darrouzet-nardi.net/downloads/dryoutcover.Rdata
>  >
>  > The code that generates my graphic almost correctly is as follows:
>  >
>  > xyplot(coversage ~ dryout | year,
>  >         data=dryoutcover,
>  >         panel = function(x,y) {
>  >                 panel.lmline(x,y)
>  >                 one <- dryoutcover$block==1
>  >                 two <- dryoutcover$block==2
>  >                 thr <- dryoutcover$block==3
>  >                 fou <- dryoutcover$block==4
>  >                 fiv <- dryoutcover$block==5
>  >                 six <- dryoutcover$block==6
>  >                 grid.points(x[one], y[one], pch=49)
>  >                 grid.points(x[two], y[two], pch=50)
>  >                 grid.points(x[thr], y[thr], pch=51)
>  >                 grid.points(x[fou], y[fou], pch=52)
>  >                 grid.points(x[fiv], y[fiv], pch=53)
>  >                 grid.points(x[six], y[six], pch=54)
>  >                 }
>  >         )
>  >
>  > The only thing wrong is that this does not correctly encode blocks 5
>  > and 6, which only appear in the third year of the study, 2003 (the
>  > third panel of the graphic). It instead labels blocks 5 and 6 as
>  > blocks 1 and 2 as if another year had started over (but on the same
>  > panel). For example, the point farthest to the right in the third
>  > panel says "2" when I want it to say "6." When I do not condition by
>  > year, it correctly displays blocks 5 and 6.
>  >
>  > How can I correct this in the conditioned graphic?
>
>This probably happens because things like `one <- dryoutcover$block==1`
>are logical vectors as long as the total number of rows in the data
>frame, whereas x and y (which you subset using x[one], y[one]) are
>shorter (only those rows for a particular year). Consequently, the
>vector you are indexing and the indexing vector are not comparable.
>
>I would suggest you use the feature meant for this sort of display,
>namely as:
>
>xyplot(coversage ~ dryout | year,
>        data=dryoutcover, groups = block, pch = c(49:54))
>
>Hope that helps,
>
>Deepayan


Oh yes, that does indeed help. I should have thought of using 
"groups". And more importantly, thanks for explaining why the 
indexing method was not working.

I have a followup question. Suppose I want to encode two different 
variables within a panel: one variable encoded by plotting character 
and one variable encoded by symbol color (as if I could use two 
"groups" variables). The dataframe I discussed above also includes a 
variable called treatment. If I start with the existing code modified 
with your suggestions:

xyplot(coversage ~ dryout | year,
	data = dryoutcover,
	groups = block,
	panel = function(x,y, ...) {
		panel.lmline(x,y)
		panel.superpose(x,y,
			pch = 49:54,
			cex = rep(2,6),
			col = rep("black", 6), ...)
		}
	)

how could I make all of the symbols of one treatment red and all of 
the symbols of the other black while maintaining the encodings of 
block by plotting character? This would be a superbly useful 
technique as it would allow 4 dimensional data on a single panel 
(maybe even 5 using a point cloud!).

Anthony