37

I have a few signals sampled over time which I plot in R using continuous lines. Additionally, I would like to highlight several specific time ranges on the plot.

My current approach is to draw full-height transparent rectangles with appropriate width and horizontal position which match the time range. I think this is a good representation, as it clearly separates points inside the range from those outside of it; but are there better ones?

And the second, practical part of the question. Now I'm plotting the signals like this:

p <- ggplot(data=gs, mapping=aes(x=frameno, y=value, col=variable)) +
       geom_line()
p

I have tried to draw an alpha blended rectangle manually:

p + geom_rect(aes(xmin=600, xmax=650, ymin=-3, ymax=3),
              colour=alpha("grey20", 0.5), fill.alpha=0.5)

— but no success so far. Any hints?

Andy
  • 4,549
  • 31
  • 26
ulidtko
  • 14,740
  • 10
  • 56
  • 88

1 Answers1

59

I think drawing rectangles just work fine, I have no idea about better solution, if a simple vertical line or lines are not enough.

And just use alpha=0.5 instead of fill.alpha=0.5 for the transparency issue also specifying inherit.aes = FALSE in geom_rect(). E.g. making a plot from the diamonds data:

p <- ggplot(diamonds, aes(x=price, y=carat)) +
     geom_line(aes(color=color))

rect <- data.frame(xmin=5000, xmax=10000, ymin=-Inf, ymax=Inf)
p + geom_rect(data=rect, aes(xmin=xmin, xmax=xmax, ymin=ymin, ymax=ymax),
              color="grey20",
              alpha=0.5,
              inherit.aes = FALSE)

alt text

Also note that ymin and ymax could be set to -Inf and Inf with ease.

Andy
  • 4,549
  • 31
  • 26
daroczig
  • 28,004
  • 7
  • 90
  • 124
  • 1
    I have tried `alpha=0.5`, that doesn't work too. It draws solid rectangle, with not transparent fill. `:(` – ulidtko Jan 19 '11 at 09:36
  • @ulidtko: you are right, specifying that a new data.frame is in use was left out. I added a concrete example, I hope you need something like that. – daroczig Jan 19 '11 at 09:58
  • 1
    Not ggplot2, but the zoo package has an `xblocks` function for this, e.g. `library(zoo); example(xblocks)` and there is also `panel.xblocks` in the latticeExtra package. – G. Grothendieck Jan 19 '11 at 12:25
  • 11
    using `..., ymin=-Inf, ymax=Inf` is absolutely brilliant. I've been struggling to get the full height of the plot covered with rectangles where ymin and ymax were "large enough", but ggplot would adjust the y scale to those values. Then I tried ylim, then with `coord_cartesian()` etc. etc. -- was going nuts! – Pierre D Jul 10 '13 at 04:28