[R] forcing levelplot to use relative cuts (ie cuts for each panel)

[Deepayan Sarkar]

On 9/13/06, Mike Townsley <uctcmkt at ucl.ac.uk> wrote:
> Dear guRus,
>
> I'm having trouble producing a levelplot with relative cuts for each
> panel (my data has large differences in scales, so I want to use
> quantiles for each panel).
>
> My attempts to change the 'at'  argument in panel.levelplot function
> have not met with success.
>
> Below is a toy example.
>
> xy <- expand.grid(x = 1:3, y = 1:3)
>
> aaa <- rbind(cbind(xy, z = 1:9, site = rep('A', 9)),
>               cbind(xy, z = (1:9)/10, site = rep('B', 9)),
>               cbind(xy, z = (1:9)*10, site = rep('C', 9)))
>
> aaa
>
> library(lattice)
> levelplot(z~x+y|site, data = aaa)         # using absolute cuts
>
> # now, attempt relative cuts
>
> levelplot(z~x+y|site, data = aaa, panel = function(...) {
>            panel.levelplot(at = quantile(z),...) })
>
> I get the following message:
> Error in panel.levelplot(at = quantile(z), ...) :
>          formal argument "at" matched by multiple actual arguments
>
> My idea was to determine the cut points each time the panel function
> is called (ie each subset of the data), but I guess this was the
> wrong thing to do.  Can someone point out what I'm missing?

Mostly that you have to catch the arguments you want to use/replace, e.g.

levelplot(z~x+y|site, data = aaa, panel = function(..., z, at) {
          panel.levelplot(..., z = z, at = quantile(z)) })

This won't actually give you want, you will need:

levelplot(z~x+y|site, data = aaa, panel = function(..., z, subscripts, at) {
          panel.levelplot(..., z = z, subscripts = subscripts, at =
quantile(z[subscripts])) })

?panel.levelplot should explain why.

Deepayan