18

I have a dataset called "merged", which contains 3 numeric columns "pauseMedian" and "numTotalPauses" and "diff". I also have a splineHull dataset, which also contains numeric columns "pauseMedian" and "numTotalPauses", plus a 6-level factor "microstyle"

I have the following code, which works perfectly. It plots a scatter plop and then overlay it with splineHull polygons colored according to "microstyle".

script 1:

ggplot(data=merged,aes(x = pauseMedian, y = numTotalPauses)) + geom_point() + geom_polygon(data = splineHull, mapping=aes(x=pauseMedian, y=numTotalPauses, group=microstyle, color = microstyle), alpha=0)

Then, I also want to change the color of the points in the scatter plot by adding just one attribute color = diff.

script 2:

ggplot(data=merged,aes(x = pauseMedian, y = numTotalPauses, color = diff)) + geom_point() + geom_polygon(data = splineHull, mapping=aes(x=pauseMedian, y=numTotalPauses, group=microstyle, color = microstyle), alpha=0)

I see the following error:

Error: Discrete value supplied to continuous scale

I don't know why I see this error. If I still want colored scatter plot but no polygons, I run the following code it works again.

script 3:

ggplot(data=merged,aes(x = pauseMedian, y = numTotalPauses, color = diff)) + geom_point()

So, what happened with script 2, where is the error from, and how can I make it work?

Improve this question
6
  • That does seem weird - it's hard to pinpoint the error without seeing example data. Also, have you tried moving color=diff to geom_point(aes(color=diff))? Commented Sep 19, 2014 at 15:12
  • @SeñorO Hi, yes I tried that. It gave the same error. In script 2, there are two color attributes, one in the ggplot aes, the other in the geom_polygon aes. The former is assigned a numeric value "diff", the latter is assigned a factor value "microstyle". I guess, maybe ggplot cannot handle a numeric color and a factor color at the same time? Commented Sep 19, 2014 at 15:16
  • Actually that may be correct now that I think about it - because it needs to make a legend for the color. Try using fill = microstyle for polygon Commented Sep 19, 2014 at 15:20
  • You don't need to post your whole dataset. Just post a small sample that will allow us to reproduce the problem using your data. For example, post of the output of dput(merged[sample(1:nrow(merged),20),]). That will give 20 randomly selected rows of your data (do the same for splineHull). Commented Sep 19, 2014 at 15:25
  • Using some fake data that (I hope) is similar to your data, I was able to get the same error message. However, when I reverse the order of geom_polygon and geom_point, I get Error: Continuous value supplied to discrete scale. There seems to be a conflict between the color scales of the two geoms, one being discrete, the other being continuous, but I'm not sure why that's happening. I would have thought having two separate geoms would result in two separate colour scales. Commented Sep 19, 2014 at 15:28

3 Answers 3

Reset to default
19

Evidently, you can't have different color aesthetics for two different geoms. As a workaround, use a fill aesthetic for the points instead. This means you have to use a point marker style that has a filled interior (see ?pch and scroll down for the available point styles). Here's a way to do that:

ggplot() + geom_point(data=merged,aes(x = pauseMedian, y = numTotalPauses, fill = diff), pch=21, size=5, colour=NA) + geom_polygon(data = splineHull, mapping=aes(x=pauseMedian, y=numTotalPauses, colour = microstyle), alpha=0)

Adding colour=NA (outside of aes()), gets rid of the default black border around the point markers. If you want a colored border around the points, just change colour=NA to whatever colour you prefer.

Also see this thread from the ggplot2 Google group, discussing a similar problem and some workarounds.

Improve this answer
Sign up to request clarification or add additional context in comments.

1 Comment

This is great (+1), but as per Juan's answer below colour = "NA" must have quotation marks, otherwise the points will not be plotted.
4

Now that we know the two color vars are of different types, there's the issue. You can try using a different scale for one (e.g. fill instead of color)

set.seed(123) my_df1 <- data.frame(a=rnorm(100), b=runif(100), c=rep(1:10, 10)) my_df2 <- data.frame(a=rnorm(100), b=runif(100), c=factor(rep(LETTERS[1:5], 20))) # this won't work. can't assign discrete and continuous to same scale ggplot() + geom_point(data=my_df1, aes(x=a, y=b, color=c)) + geom_polygon(data=my_df2, aes(x=a, y=b, color=c), alpha=0.5)

Error: Discrete value supplied to continuous scale

# but use fill for polygons, and that works: ggplot() + geom_point(data=my_df1, aes(x=a, y=b, color=c)) + geom_polygon(data=my_df2, aes(x=a, y=b, fill=c), alpha=0.5)

plot output

If you have to use the same scale (color), and can't convert the variables to the same type, see this for more info: Plotting continuous and discrete series in ggplot with facet

Improve this answer

3 Comments

Hi, I've got a very large dataset, so it's difficult to post the data here. alpha should be more than 0, but it is not the problem, and I only need to contour not filled polygon, because the plot simply does not show up due to the error. I tried your solution but I got the same error. For the color specified in geom_point, it is a numeric variable "diff", the one specified in geom_polygon is a discrete "factor". I am just guessing ggplot does not work if color attributes are assigned with different types.
Oh, well that's the key piece of missing info. See new solution
Hi, thanks. Yes it works this way. Actually I still prefer color, because I want to have just the contour (empty polygons). Anyway, this solution works for now. Thanks!
2

Just to add something to the preferred eipi10 answer above (thank you for that!!). The colour="NA" option in order to get rid of the border around the circle shape (pch=21) has to be between "". If you use colour=NA (without the quotation marks) the whole shape disappears and is not plotted. I would have just commented on the answer but I still don't have reputation for that :)

Improve this answer

1 Comment

So glad to see this! I spent a long time wondering why I couldn't get it to work. Although colour = "NA" will give a white border, so if there are overlapping points/ geoms the border will still be visible.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.