1

I am trying to get a plot (using ggplot) that will plot geom_point and geom_line where different variables are plotted in different colors (according to scale_colour_manual(value = color) where color is a custom array of colors) in addition to two horizontal black lines (geom_hline).

My trouble arises when I attempt to get the legend text for the horizontal lines customized. It appears that I can have only one of the two:

  1. the color of the horizontal line as black but the legend text for this line is incorrect
  2. the legend text for this line is correct but the color of the horizontal line is determined by the aforementioned scale_colour_manual color.

plot <- ggplot(data, aes(x = factor(Month), y = avgLoss, colour = type, order = -as.numeric(type)))
         + geom_line() + geom_point()

meanplus2sd <- mean(data$avgLoss) + 2*sd(data$avgLoss)

plot <- plot + geom_hline(aes(yintercept = meanplus2sd), colour = "black")

produces black line that says "black" in the legend

plot <- plot + geom_hline(aes(yintercept = meanplus2sd, colour = "Mean + 2 Stdev."))

produces a line that is the next color in my defined scale_colour_manual array, but the legend text is "Mean + 2 Stdev."

Any help in getting both custom color and legend text for a horizontal line in addition to the standard geom_point + geom_line plotting would be excellent. Thanks.

Marek
  • 49,472
  • 15
  • 99
  • 121
Peaverton
  • 9
  • 1
  • 3
  • A reproducible example (ie some code) will greatly improve the chances that someone can answer this question. – joran May 18 '11 at 20:22
  • 2
    Follow the guidelines here: http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example to give a reproducible example. This question should not be difficult, but the answer will depend on what code you've already gotten written. – Chase May 18 '11 at 20:25
  • plot <- ggplot(data, aes(x = factor(Month), y = avgLoss, colour = type, order = -as.numeric(type))) + geom_line() + geom_point() meanplus2sd <- mean(data$avgLoss) + 2*sd(data$avgLoss) plot <- plot + geom_hline(aes(yintercept = meanplus2sd), colour = "black") ... produces black line that says "black" in the legend plot <- plot + geom_hline(aes(yintercept = meanplus2sd, colour = "Mean + 2 Stdev.")) ... produces a line that is the next color in my defined scale_colour_manual array, but the legend text is "Mean + 2 Stdev." – Peaverton May 18 '11 at 20:30
  • One last bit - how about an example of `data` so that others can copy / paste your code exactly as you have it. You can use `head(dput(data), 20)` and then paste the results of that back into the question. Alternatively, make up some sample data using `sample`, `rnorm`, et al. – Chase May 18 '11 at 20:48

2 Answers2

1

One way to get what you want would be to include the data for your horizontal black line in the original data frame. You example isn't exactly reproducible, so this may not be exactly the code you want, but this is the general idea:

Add rows to 'data', one for each level of 'Month', where the value of 'avgLoss' in each row is equal to 'meanplus2sd'. You'll probably want all the other columns to contain NAs. So,

newData <- head(data,length(unique(data$Month)))
newData$Month <- unique(data$Month) 
newData$avgLoss <- rep(meanplus2sd,length(unique(data$Month)
newData$type <- rep('Mean + 2sd',length(unique(data$Month)))
#Then set all other values in newData to NA...and,
data <- rbind(data,newData)

That should get you started, you'll just have to make sure that the color "black" is in the right position in scale_colour_manual, which may take some fiddling.

joran
  • 169,992
  • 32
  • 429
  • 468
  • Not a bad work around, except I'll end up with points in addition to the line. I also thought about having the line plotted prior to the geom_point which should then make the horizontal line have the first color in the array...then it's just about changing the color array for this function. I was/am hoping there is an explicit way to do this that isn't a work-around, though. – Peaverton May 18 '11 at 21:11
  • 1
    No necessarily, I think, if you pass the appropriate subsets of your data to geom_point and geom_line. Pass all the data to geom_line, but omit the horizontal line data in geom_point. – joran May 18 '11 at 21:44
0

Annotate might be a good option

x <- baseball
tmp <- subset(x, team == c("CHN","NYA"))
cols <- c("8" = "red","4" = "blue","6" = "darkgreen")
meanplus2sd <- mean(tmp$rbi) + 2*sd(tmp$rbi)

plot <- ggplot(tmp, aes(x = year, y = rbi, color = factor(team), order = -as.numeric(team))) + 
geom_line() + geom_point() + geom_hline(aes(yintercept = meanplus2sd)) + 
annotate("text",label="Mean + 2sd",x=min(tmp$year)+10, y=meanplus2sd+10)
DJJ
  • 2,481
  • 2
  • 28
  • 53
Brandon Bertelsen
  • 43,807
  • 34
  • 160
  • 255