R - xyplot (lattice); multiple response (Y axis) - how do I identify individual points?

Question

I've generated a lattice plot with 2 response variables plotted on a single Y axis). I have plotted one of the response variables as a line (Response1 below), the other (Response2) as points on that same Y axis. I wish to be able to highlight some of the points (in Response 2) according to a condition ('Condition'), for example, if Condition==1 then change the point color to blue and increase the point size and change the symbol. The fact that I wish to plot two response variables on the same axis excludes the 'Group' option for this and, consequently, the panel.superpose function (as far as i can tell). Example code is given below:

library(lattice)
#generate dataframe
TD=data.frame(Response1=rnorm(100,50,5),Response2=rnorm(100,70,5),
Xaxis=seq(1:100),Factor=rep(LETTERS[1:5],20),
Condition=sample(0:1,100,replace=T))
#generate plot          
xyplot(Response1+Response2~Xaxis|Factor,data=TD,distribute.type=TRUE,
type=c('l','p'))
#if(condition==1) then Response2 symbol=blue and size is larger??

I have tried to do this via latticeExtra's doubleYScale plot (using a 'group' argument' in one of the plots to condition on 'Condition') but this is lost when the overlay is made.

Any guidance on this would be much appreciated.

Regards

Tom.

Answer 1

Manipulating an input dataset is not always desirable and might render your code unnecessarily confusing. Instead, you could create two separate xyplot objects - one with both response variables represented as points and lines, one with only those points of the second response variable for which Condition == 1 applies - and blend them together using as.layer from latticeExtra.

library(latticeExtra)

## grouped plot with lines and points
p <- xyplot(Response1 + Response2 ~ Xaxis | Factor, data = TD, 
            distribute.type = TRUE, type = c("l", "p"), as.table = TRUE) 

## plot with 'Condition == 1' points only
p_cond1 <- xyplot(Response2 ~ Xaxis | Factor, 
                  data = subset(TD, Condition == 1), col = "green")

## combine plots
p + as.layer(p_cond1)

Similarly, if you would like to modify the point symbol and size, I would recommend creating three different xyplot objects - one with the first response variable represented as lines, one with only those points of the second response variable for which Condition == 0 applies, and similarly one for Condition == 1.

## grouped plot with lines only
p <- xyplot(Response1 + Response2 ~ Xaxis | Factor, data = TD, 
            distribute.type = TRUE, type = c("l", "p"), as.table = TRUE, 
            col.symbol = "transparent") 

## plot with 'Condition == 0' (or 'Condition == 1') points only
p_cond <- lapply(0:1, function(i) {
  xyplot(Response2 ~ Xaxis | Factor, 
         data = subset(TD, Condition == i), 
         col = ifelse(i == 0, "orange", "darkgreen"), 
         pch = ifelse(i == 0, 1, 2), cex = ifelse(i == 0, 0.8, 1.2))
})

## combine plots
p + 
  as.layer(p_cond[[1]]) + 
  as.layer(p_cond[[2]])

Answer 2

You can split Response2 into separate variables according to Condition. Here is one way.

TD = within(TD, {
  R2a=Response2*Condition
  R2b = Response2*(1-Condition)
  R2a[R2a==0]=NA
  R2b[R2b==0]=NA
})

xyplot(Response1+R2a+R2b~Xaxis|Factor,data=TD,distribute.type=TRUE,
type=c('l','p','p'))

Then adjust size and color to suit you.