stat_count does not work with line

[dpmcsuss]

When I make an empirical probability mass functions (pmf), I like to use linerange with ymin=0 and ymax equal to the value of the empirical pmf. stat_count works well but I cannot use linerange with it without using kind of weird hack.

library(ggplot2)
# works and gives a bar chart
ggplot(mtcars,aes(x=gear))+stat_count()

# throws an error
ggplot(mtcars,aes(x=gear))+stat_count(aes(ymax=..prop..,ymin=0),geom="linerange")
# Error: stat_count() must not be used with a y aesthetic.

# this works because it does not populate the ymin variable
ggplot(mtcars,aes(x=gear))+stat_count(aes(ymax=..prop..,ymin=..prop..-..prop..),geom="linerange")

The reason that it fails is because when I set ymin=0 in aes, the data variable has a column with name ymin and is.null(data$y) returns true which results in the error. The hack involves setting ymin=..prop..-..prop.. in the aes call.

ggplot2/R/stat-count.r

Line 53 in c592e32

if (!is.null(data$y)) {

[has2k1]

How about?

ggplot(mtcars,aes(x=gear))+stat_count(width=0.005)

For one, the desirable width will depend on the x scale. It is less wierd though.

[dpmcsuss]

Yes there are definitely some workarounds and not sure if this is really a bug but it is somewhat weird (though very R) behavior. To me, the width solution is less appealing than theymin=..prop..-..prop.. solution.

[hadley]

I think this will be fixed by #2059, since it would allow you to write:

ggplot(mtcars, aes(gear)) + 
  stat_count(aes(ymax = computed(prop), ymin = 0, y = computed(0)), geom="linerange")

Otherwise, given that there are other ways to solve the problem (i.e. counting yourself) and it affects few people, I unfortunately don't have the time to make a fix.

[dpmcsuss]

Thanks. That looks like it will be a nice solution and I definitely understand that it is a minor issue.