4

I have written an answer here and would like to improve it. What I would like to do is to remove the legend for geom_path but it is not working with show.legend = FALSE. The color elements of geom_path remain in the legend (Down and Up). Am I doing something wrong?

Is there alternatively a way to manually tell ggplot just to show lets say the last two elements of the legend (y2015, y2016)?

My Code and the Output: 

library(ggplot2)
library(reshape2)
library(dplyr)

ggplot2df <- read.table(text = "question y2015 y2016
                        q1 90 50
                        q2 80 60
                        q3 70 90
                        q4 90 60
                        q5 30 20", header = TRUE)


df <- ggplot2df %>% 
  mutate(direction = ifelse(y2016 - y2015 > 0, "Up", "Down"))%>%
  melt(id = c("question", "direction"))


ggplot(df, aes(x=question, y = value, color = variable, group = question )) + 
geom_point(size=4) + 
geom_path(aes(color = direction), arrow=arrow(), show.legend = FALSE)

enter image description here

Community
  • 1
  • 1
Alex
  • 4,925
  • 2
  • 32
  • 48

2 Answers2

3

I think what's going on is that because both variable and direction are mapped to color, the legend has four different color values. Removing the path legend just removes the arrows, while removing just the point legend removes the points. But either way, all four colors still show up in the legend, as points or arrows, respectively, because the underlying mapping still has four values, regardless of whether you choose to manifest those four values as points, arrows, or both in the legend.

One way around this would be to use a fill aesthetic for the points. Then the path legend will have only two values. To do this, you have to use a point style with a filled interior (pch values 21 - 25). You'll also need to change the colors of either the color aesthetic or fill aesthetic, so that they won't be the same:

ggplot(df, aes(x=question, y = value, group = question)) + 
  geom_point(size=4, aes(fill=variable), pch=21, color=NA) + 
  geom_path(aes(color = direction), arrow=arrow(), show.legend=FALSE) +
  scale_fill_manual(values=hcl(c(105,285), 100, 50))

enter image description here

eipi10
  • 91,525
  • 24
  • 209
  • 285
  • Thanks (and +1)! That is a nice workaround. Could you move `show.legend=FALSE` to `geom_path` in order to match the question. I will however not directly accept because I have the feeling that there must be something more and hence would like to encourage other answers. – Alex Jun 29 '16 at 21:13
  • Yes, sorry. Got those backward. – eipi10 Jun 29 '16 at 21:15
  • I actually think that in this particular case there are better ways to encode the information ([see my answer to SO question you linked to](http://stackoverflow.com/a/38108908/496488)), but using a fill aesthetic for the points is still a nice trick to have handy when you'd like create the equivalent of two separate color aesthetics. – eipi10 Jun 29 '16 at 21:26
  • I will look into it. Thanks! – Alex Jun 29 '16 at 21:33
2

Another possibility (but more overall work) is to manually specify the colors for the arrows:

library(ggplot2)
library(tidyr)
library(dplyr)

ggplot2df <- read.table(text = "question y2015 y2016
q1 90 50
q2 80 60
q3 70 90
q4 90 60
q5 30 20", header = TRUE)

ggplot2df %>% 
  mutate(direction = ifelse(y2016 - y2015 > 0, "Up", "Down")) %>%
  gather(variable, value, -question, -direction) -> df

gg <- ggplot(df, aes(x=question, y = value, group = question)) 
gg <- gg + geom_point(aes(color=variable), size=4) 
gg <- gg + geom_path(color=c("red", "red", "green", rep("red", 4), "green", "red", "red"),
                     arrow=arrow(), show.legend=FALSE)  
gg

enter image description here

hrbrmstr
  • 77,368
  • 11
  • 139
  • 205