Question

I have data conditioned on two variables, one major condition, one minor condition. I want a xyplot (lattice) with points and lines (type='b'), in one panel so that the major condition determines the color and the minor condition is used for drawing the lines.

Here is an example that is representative of my problem (see the code below to produce the data frame). d is the major condition, and c is the minor condition.

> dat
     x          y c d
1    1  0.9645269 a A
2    2  1.4892217 a A
3    3  1.4848654 a A
....
10  10  2.4802803 a A
11   1  1.5606218 b A
12   2  1.5346806 b A
....
98   8  2.0381943 j B
99   9  2.0826099 j B
100 10  2.2799917 j B

The way to get the connecting lines to be conditioned on c is to use groups=c in the plot. Then the way to tell them apart is to use a formula conditioned on d:

xyplot(y~x|d, data=dat, type='b', groups=c)

However, I want the plots in the same panel. Removing the formula condition on d produces one panel, but when group=d is specified, there are “retrace” lines drawn:

xyplot(y~x, data=dat, type='b', groups=d, auto.key=list(space='inside'))

What I want looks very like the above plot, only without these “retrace” lines.

It’s possible to set the colors explicitly in this example, as I know that there are five lines of category ‘A’ followed by five of category ‘B’, but this won’t easily work for my real problem. In addition, auto.key is useless when setting the colors this way:

xyplot(y~x, data=dat, type='b', groups=c, col=rep(5:6, each=5))

The data:

set.seed(1)
dat <- do.call(
  rbind,
  lapply(1:10,
         function(x) {
           firsthalf <- x < 6
           data.frame(x=1:10, y=log(1:10 + rnorm(10, .25) + 2 * firsthalf),
                                c=letters[x],
                                d=LETTERS[2-firsthalf]
                                )
         }
  )
)

Answer 1

The default graphical parameters are obtained from the superpose.symbol and superpose.line. One solution s to set them using par.settings argument.

## I compute the color by group 
col <-by(dat,dat$c,
                FUN=function(x){
                  v <- ifelse(x$d=='A','darkgreen','orange')
                  v[1]  ## I return one parameter , since I need one color
                  }
         )

 xyplot(y~x, data=dat, type='b', groups=c,
       auto.key = list(text =levels(dat$d),points=F),
       par.settings=
         list(superpose.line   = list(col = col),                 ## color of lines
              superpose.symbol = list(col=col),                   ## colors of points
              add.text = list(col=c('darkgreen','orange'))))      ## color of text in the legend