inherit mapping in geom_miss_point

[jminnier]

Hey there, I love your package and especially the geom_miss_point function, but I think I found some unexpected behavior. My issue is that I cannot seem to get geom_miss_point to inherit the fill or shape in the aes(). Here's a reprex showing what I mean:

library(ggplot2)
library(naniar)

mydat <- mtcars
mydat$mpg[1:10] = NA

Here is a simple missing plot that I'm using as a basis:

ggplot(mydat,aes(x=mpg,y=disp))+geom_point()+geom_miss_point()
#> Warning: Removed 10 rows containing missing values (geom_point).

Example showing it works for regular geom_point: fill and color are inherited with geom_point just fine and this is what I'd like to emulate in geom_miss_point.

ggplot(mtcars,aes(x=mpg,y=disp,fill=cyl))+
  geom_point(size=4,alpha=0.5,pch=21)+
  geom_point(aes(x=mpg+2),pch=21)

ggplot(mtcars,aes(x=mpg,y=disp,color=cyl))+
  geom_point(size=4,alpha=0.5)+
  geom_point(aes(x=mpg+2))

However, fill is not inherited with geom_miss_point. I'd like to have the missing points filled with their respective values of cyl:

ggplot(mydat,aes(x=mpg,y=disp,fill=cyl))+
  geom_point(pch=21,size=4)+
  #geom_point(pch=21,aes(x=mpg+2))+
  geom_miss_point(pch=21,size=4)
#> Warning: Removed 10 rows containing missing values (geom_point).

also, trying to inherit size gives an error:

ggplot(mydat,aes(x=mpg,y=disp,size=cyl))+
  geom_point(pch=21)+
  geom_miss_point(pch=21)
#> Warning: Removed 10 rows containing missing values (geom_point).
#> Error in rep(value[[k]], length.out = n): attempt to replicate an object of type 'symbol'

Specifying fill in the function does not work, either

ggplot(mydat,aes(x=mpg,y=disp,fill=cyl))+geom_point(pch=21,size=4)+
  geom_miss_point(aes(fill=cyl),pch=21,size=4)
#> Warning: Removed 10 rows containing missing values (geom_point).

But specifying fill as a constant does work.

ggplot(mydat,aes(x=mpg,y=disp,fill=cyl))+geom_point(pch=21,size=4)+
  geom_miss_point(fill="red",pch=21,size=4)
#> Warning: Removed 10 rows containing missing values (geom_point).

I tried forking your repo and playing around with geom_miss_point and GeomMissPoint but I couldn't figure out how to fix it. I'm not very experienced with extending ggplot2 and I think this is beyond my knowledge. I also tried updating ggplot2 to the development version but that didn't make a difference. Wish I could have been more helpful in figuring this out! I will keep trying, but perhaps if you get a chance to check it out it will be easy for you to see what's causing this behavior?

Thanks for naniar, it truly is absurdly useful!

[seasmith]

Using "miss_point" within geom_point() gives:

ggplot(mydat) +
    geom_point(aes(x=mpg,y=disp, color = ..missing.., fill= cyl),
               pch = 21, size = 4, stat = "miss_point")

So I made a slight mod to StatMissPoint:

ggplot(mydat) +
    geom_point(aes(x=mpg,y=disp, color = ..missing.., fill= cyl),
               pch = 21, size = 4, color = "black",stat = "miss_point")

The mod I made in StatMissPoint was to compute_group:

data.frame(data,
           missing = missing_label)

However, I did notice this behavior:

# Works using the unmodified `StatMissPoint`
ggplot(mydat, aes(x = mpg, y = disp, fill = factor(cyl)))+
    geom_point(pch = 21, size = 4)+
    stat_miss_point(pch = 21, size = 4)

# Throws the same error in both cases of `StatMissPoint`
ggplot(mydat, aes(x = mpg, y = disp, fill = factor(cyl)))+
    geom_point(pch = 21, size = 4)+
    geom_miss_point(pch = 21, size = 4)

#> Error in rep(value[[k]], length.out = n) : 
#>  attempt to replicate an object of type 'symbol'
#> In addition: Warning message:
#> Removed 10 rows containing missing values (geom_point). 

[njtierney]

Hi!

Thanks for filing the complement, and the bug report!

Ideally geom_miss_point() should behave as geom_point .

@seasmith - can you point to the changes you made in StatMissingPoint ?

This should be fixed in the next release.

[njtierney]

Also, thank you to both of you @jminnier and @seasmith for trying to work out how to implement this - really appreciate it! :)

[njtierney]

Thanks to the kind work of @seasmith I think that this works now:

library(ggplot2)
library(naniar)

ggplot(airquality,
       aes(x = Ozone,
           y = Solar.R,
           colour = factor(Month))) +
  geom_miss_point()

Created on 2018-04-17 by the reprex package (v0.2.0).

Let me know how you go, @jminnier ? And thank you so much for filing an issue!

[jminnier]

Ah, awesome, thanks so much @seasmith! It works perfectly, now. Keep up the great work, both of you!