How to enforce ggplot's position_dodge on categories with no data?

20
votes

I'm trying to use position_dodge on ggplot to obtain boxplots of two different signals (ind) sharing the same categories (cat). When there is a category with data for one signal but not for the other one, the boxplot for the signal with data covers all the horizontal spacing, and does not respect the position_dodge instruction for that particular category. Is there a way to make ggplot to enforce the dodging rule? As you can see on the example below, the signal x has no data for category B, so it loses the space reserved by position_dodge. I would like to avoid that.

Thanks in advance.

data<-data.frame(cat=c('A','A','A','A','B','B','A','A','A','A','B','B'), 
                 values=c(3,2,1,4,NA,NA,4,5,6,7,8,9), 
                 ind=c('x','x','x','x','x','x','y','y','y','y','y','y'))

print(ggplot() +
        scale_colour_hue(guide='none') +
      geom_boxplot(
           aes(x=as.factor(cat), y=values, 
               fill=ind), 
           position=position_dodge(width=.60), 
           data=data,
           outlier.size = 1.2,
           na.rm=T))

graph with original problem

PROGRESS UPDATE

After some workarounds, I came up with the outcome I was looking for... (kind of) 

data            <- data.frame(
cat=c('A','A','A','A','B','B','A','A','A','A','B','B','B'), 
values=c(3,2,1,4,NA,NA,4,5,6,7,8,9, 0), 
ind=c('x','x','x','x','x','x','y','y','y','y','y','y','x'))

p  <- ggplot() +
      scale_colour_hue(guide='none') +
      geom_boxplot(aes(x=as.factor(cat), y=values, fill=ind),
      position=position_dodge(width=.60), 
      data=data,
      outlier.size = 1.2,
      na.rm=T) +
      geom_line(aes(x=x, y=y), 
                data=data.frame(x=c(0,3),y=rep(0,2)), 
                size = 1, 
                col='white')
print(p)

solution with workaround

Some people remcomended using faceting for the effect I wanted. Faceting doesn't give me the effect I'm looking for. The final graph I was looking for is shown below:

final graph

If you notice, the white major tick mark at y = 10 is thicker than the other tick marks. This thicker line is the geom_line with size=1 that hides unwanted boxplots.

I wish we could combine different geom objects more seamlessly. I reported this as a bug on Hadley's github, but Hadley said this is how position_dodge behaves by design. I guess I'm using ggplot2 in a non-standard way and workarounds are the way to go on these kind of issues. Anyways, I hope this helps some of the R folks to push ggplot great functionality a little further.

3
rggplot2
Is there any better way to handle this 2.5 years later? I have the same situation. I have tried using expand.grid() to create a data set with NA's for all the missing factor level combinations of the variable I am plotting, and they seem to just be ignored. – Michael
github.com/hadley/ggplot2/issues/688 Maybe not :( – Michael
I which there was. I wrapped the ggplot code into a function and made the white line y-axis position a parameter of this function. The reason is, if data crosses y=0, it will fall underneath the white line, making the workaround too obvious and distracting. – JAponte
This is the best solution I found. stackoverflow.com/questions/15367762/… (The checked answer, not the one where you make fake data.) – Michael
Interesting. I wonder how this would work for an arbitrary number of X values and missing values. I bet there is a programmatic workaround. I like the fact that it doesn't rely on dummy values and white lines. – JAponte

3 Answers

2
votes

x of B has no values, so you can add "B", 0, "x" which essentially indicates that there is no distribution of "values" for x of B. The median and other percentiles are zero. 

 data<-data.frame(cat=c('A','A','A','A','B','B','A','A','A','A','B','B','B'), 
             values=c(3,2,1,4,NA,NA,4,5,6,7,8,9,0), 
             ind=c('x','x','x','x','x','x','y','y','y','y','y','y','x'))

Also you do not have to add position parameters here, because when you consider x as a factor, ggplot -- geom_boxplot will automagically dodge to the sides.

print(ggplot() +
  scale_colour_hue(guide='none') +
  geom_boxplot(aes(x=as.factor(cat), y=values, fill=ind), 
  data=data,
  outlier.size = 1.2,
  na.rm=T))

2
votes

I just got a clue to use faceting from one of the comments posted by Hadley at his git site, so credits goes to Hadley, the maintainer of ggplot2 package!

See if this is what you wanted. To learn more about options on setting the whiskers and others in this plot, check this help page in ggplot2 package:

?stat_boxplot

data<-data.frame(cat=c('A','A','A','A','B','B','A','A','A','A','B','B'), 
             values=c(3,2,1,4,NA,NA,4,5,6,7,8,9), 
             ind=c('x','x','x','x','x','x','y','y','y','y','y','y'))

p <- ggplot(data = data, aes(factor(cat), values))                     
p + stat_boxplot(geom="boxplot", position = "dodge", width = 0.60, na.rm = TRUE) +  facet_grid(.~ind)

enter image description here

To add colors to your plot, which in my opinion is a redundant one as you are already faceting the plot based on "ind" variable, try this:

p <- ggplot(data, aes(factor(cat), values, fill = ind))                     
p + stat_boxplot(geom="boxplot", position = "dodge", width = 0.60, na.rm = TRUE) + facet_grid(.~ind)

enter image description here

HTH!

2
votes

After some workarounds, I came up with the outcome I was looking for... (kind of) 

data            <- data.frame(
cat=c('A','A','A','A','B','B','A','A','A','A','B','B','B'), 
values=c(3,2,1,4,NA,NA,4,5,6,7,8,9, 0), 
ind=c('x','x','x','x','x','x','y','y','y','y','y','y','x'))

p  <- ggplot() +
      scale_colour_hue(guide='none') +
      geom_boxplot(aes(x=as.factor(cat), y=values, fill=ind),
      position=position_dodge(width=.60), 
      data=data,
      outlier.size = 1.2,
      na.rm=T) +
      geom_line(aes(x=x, y=y), 
                data=data.frame(x=c(0,3),y=rep(0,2)), 
                size = 1, 
                col='white')
print(p)

solution with workaround

Some people recommended using faceting for the effect I wanted. Faceting doesn't give me the effect I'm looking for. The final graph I was looking for is shown below:

final graph

If you notice, the white major tick mark at y = 10 is thicker than the other tick marks. This thicker line is the geom_line with size=1 that hides unwanted boxplots.

I wish we could combine different geom objects more seamlessly. I reported this as a bug on Hadley's github, but Hadley said this is how position_dodge behaves by design. I guess I'm using ggplot2 in a non-standard way and workarounds are the way to go on these kind of issues. Anyways, I hope this helps some of the R folks to push ggplot great functionality a little further.