[R] coloring individual points in lattice xyplot

[Gabor Grothendieck]

On 7/5/06, Deepayan Sarkar <deepayan.sarkar at gmail.com> wrote:
> On 7/4/06, Gabor Grothendieck <ggrothendieck at gmail.com> wrote:
> > OK.  It looks like I need to go to the lower level llines and lpoints to
> > do this.  I wrote a panel routine, mypanel, and it seems to work (see
> > below); however, currently it assumes types and cols are in the global
> > environment or at least somewhere where they will be found.
> >
> > 1. Is it somehow possible to stick these into some structures set up by
> > lattice already and then retrieve them from lattice from within mypanel?
>
> Not sure what you mean by that, but I would do something like the
> following (this allows 'col' to be a list as in your original post).
> Note that arguments to the panel function can be supplied directly to
> the high level function (that's why graphical parameters can be
> supplied to xyplot in the first place).
>
> mypanel <-
>    function(x, y, subscripts, groups,
>             col = 1,
>             type = "p",
>             ...)
> {
>    col <- rep(as.list(col), length = nlevels(groups))
>    type <- rep(as.list(type), length = nlevels(groups))
>    for(g in 1:nlevels(groups)) {
>         idx <- g == groups[subscripts]
>         xx <- x[idx]; yy <- y[idx];
>         panel.xyplot(x[idx], y[idx],
>                      col = col[[g]],
>                      type = type[[g]],
>                      ...)
>    }
> }
>
> xyplot(y ~ c(x,x), groups = factor(col(y)),
>       panel = mypanel,
>       type = c("o", "p"),
>       col = list("black", 1:10))


Thanks!!! This is quite an improvement.

>
>
> > 2. Any other improvements to the example below?
> >
> > mypanel <- function(x, y, subscripts, groups, ...) {
> >       for(g in 1:nlevels(groups)) {
> >          idx <- g == groups
>
> This won't work for more than one panel.

Could you explain this?  Under what situation does it not work?
The following, for example, appears to work:

xyplot(y ~ c(x,x) | col(y), groups = factor(col(y)),
      panel = mypanel,
      type = c("o", "p"),
      col = list("black", 1:10),
      layout = 1:2)




>
> >          xx <- x[idx]; yy <- y[idx]; ccols <- cols[subscripts][idx]
> >          if (any(idx)) {
> >             switch(types[g],
> >                p = lpoints(xx, yy, col = ccols),
> >                l = llines(xx, yy, col = ccols),
> >                o = { lpoints(xx, yy, col = ccols)
> >                         llines(xx, yy, col = ccols) })
> >          }
> >        }
> >     }
> >
> > x <- 1:10
> > y <- cbind(y1 = x, y2 = x+1)
> > cols <- c(rep(1,10), 1:10)
> > types <- c("o", "p")
> > xyplot(y ~ c(x,x), groups = factor(col(y)), type = types, panel = mypanel)
> >
> >
> >
> > On 7/4/06, Deepayan Sarkar <deepayan.sarkar at gmail.com> wrote:
> > > On 7/4/06, Gabor Grothendieck <ggrothendieck at gmail.com> wrote:
> > > > I can get the types to work or the colors but not both:
> > >
> > > Sorry if I wasn't clear, but I didn't mean that you could use
> > > panel.superpose[.2] to do what you wanted. I only meant that you could
> > > use it as a template that may help you to write your own panel
> > > function. What you want is not possible with tools available in
> > > lattice.
> > >
> > > Deepayan
> > >
> > > > # this gets the types right but not the colors
> > > > library(lattice)
> > > > x <- 1:10
> > > > y <- cbind(y1 = x, y2 = x+1)
> > > > cols <- c(rep(1,10), 1:10)
> > > > xyplot(y ~ c(x,x), groups = col(y), type = c("o", "p"),
> > > >    panel = function(x, y, subscripts, groups, ...)
> > > >       panel.superpose.2(x, y, subscripts, groups, col = cols[subscripts], ...)
> > > > )
> > > >
> > > >
> > > > # this gets the colors right but not the types
> > > > library(lattice)
> > > > x <- 1:10
> > > > y <- cbind(y1 = x, y2 = x+1)
> > > > cols <- c(rep(1,10), 1:10)
> > > > xyplot(y ~ c(x,x), groups = col(y), type = c("o", "p"),
> > > >    panel = function(x, y, subscripts, groups, ...)
> > > >       panel.xyplot(x, y, col = cols[subscripts], ...)
> > > > )
> > > >
> > > >
> > > > On 7/4/06, Gabor Grothendieck <ggrothendieck at gmail.com> wrote:
> > > > > On 7/4/06, Deepayan Sarkar <deepayan.sarkar at gmail.com> wrote:
> > > > > > On 7/4/06, Gabor Grothendieck <ggrothendieck at gmail.com> wrote:
> > > > > > > If I wish to color groups in xyplot I can do this:
> > > > > > >
> > > > > > >    library(lattice)
> > > > > > >    x <- 1:10
> > > > > > >    y <- cbind(x, x+1)
> > > > > > >    xyplot(y ~ rep(x,2), group = col(y), col = 1:2)
> > > > > > >
> > > > > > > How do I color different points differently within a group.
> > > > > > >
> > > > > > > For example, I want to produce this plot (except that I only
> > > > > > > want to have two groups, not 11):
> > > > > > >
> > > > > > >    xyplot(y ~ rep(x,2), group = c(rep(1, 10), 2:11), col = 1:11)
> > > > > > >
> > > > > > > I am thinking of something like this (although
> > > > > > > this does not work, its just to get the idea across):
> > > > > > >
> > > > > > >    xyplot(y ~ rep(x,2), group = col(y), col = list(1, 2:11))
> > > > > > >
> > > > > > > where, in general, I have a list with one component per group
> > > > > > > whose elements are scalars to color the whole group or
> > > > > > > vectors one color per point in the group.  I don't know
> > > > > > > ahead of time what the list is.
> > > > > > >
> > > > > > > I am looking for a general approach to this within the lattice
> > > > > > > xyplot plot framework; the above is just an example.
> > > > > >
> > > > > > The general approach is to write your own panel function. For a
> > > > > > possible template, look at the functions panel.superpose and
> > > > > > panel.superpose.2 and how they handle the 'type' argument.
> > > > > >
> > > > > > Deepayan
> > > > > >
> > > > >
> > > > > There is no example in ?panel.superpose.  Do you think you
> > > > > could provide an example for the situation in my post?
> > > > >
> > > > > I have done quite a bit of RSiteSearch'ing and googling prior to
> > > > > posting and all the examples I found had colors that depended
> > > > > on the group, none addressed the situation in my post -- i.e.
> > > > > coloring individual points within groups.
> > > > >
> > > >
> > >
> > >
> > > --
> > > http://www.stat.wisc.edu/~deepayan/
> > >
> >
>
>
> --
> http://www.stat.wisc.edu/~deepayan/
>