Is there a way to add a legend for ggplot's alpha?

Question

I have a plot with many overlapping points (goes from 2-10). Adding jitter to the points makes it very noisy and unappealing. I like adding a alpha in the aesthetics. However, I'd like to have a legend where a reader can see how many points are overlapping for each of those transparencies. Is there such a way?

ggplot(data=mydata,aes(x=x,y=y)) + geom_point(size=3,shape=2,aes(alpha=1/3))

Let's say I use the above code. How would I incorporate a legend for the alpha?

Answer 1

Here is an example of how to fake this. Try it a couple times while varying alpha .

require(ggplot2)

n = 10000
alpha = 0.01

set.seed(12345)
data = data.frame(replicate(2, rnorm(n)))

dev.new(width=4, height=3)
p = qplot(X1, X2, data=data, alpha=alpha)

fake_scale = scale_alpha('# of overlapping pts', breaks=pretty(c(alpha, 1)), labels=as.character(pretty(c(alpha, 1))/alpha))

p + fake_scale

alpha = 0.1

alpha = 0.01

Answer 2

Not exactly what you want, but how about geom_hex()?

If you don't discretize (bin) it, I think R would need to calculate the overlapped area and the number of overlappedness(sp?) (which would also depend on point size), and that sounds hard.

library(hexbin)

mydata <- data.frame(x = rnorm(100), y = rnorm(100))
ggplot(data=mydata,aes(x=x,y=y)) + geom_hex()

Answer 3

Taking cues from other answers, it seems like you may want to bin the xy points into unique values, and then sum up the number of unique values per each xy coordinate bin.

Using the fake data from another answer,

library(ggplot2)
library(dplyr)
library(magrittr)

n = 10000

set.seed(12345)
data = data.frame(replicate(2, rnorm(n)))

I'm binning by the 1st decimal place, and then counting how many are in each bin

data2 <- data %>% mutate(
  x = round(X1,1),  
  y = round(X2,1)
) %>% 
  group_by(x,y) %>%
  tally() %>% 
  dplyr::filter(n<=10) %>% 
  mutate(transp = n/10) 

I filtered out the biggest bins just for illustration purposes. Now, the transp column has the calculation you want, which you can supply to alpha and get it in the legend

data2 %>% 
  ggplot(aes(x=x,y=y)) + 
  geom_point(aes(alpha = as.factor(transp))) + 
  scale_alpha_discrete(
    "Number of overlapping points",
    labels = 1:10
  )+ 
  theme(legend.position = "bottom")

pic: