make a common legend and include unique legend attributes

Question

I am attempting to combine a legend for two plots with a shared blue line and unique bar colors (read and yellow). When plotting the code below only two of the three colored id's show up in the combined legend (red and blue) but I want the third color to show up as well in the combined legend.

For example

red as beaver temp1
yellow as beaver temp2
blue as line

Here the line is a stand in for what I am using for a moving average. I don't think facet wrapping this will work as there are additional variable parameters in my code (i.e. scale_y_continuous that have different secondary axis scales) but below is a reproducible example of the basic idea.

library(datasets)
library(ggplot2)
library(cowplot)
data(beavers)

colour <- c("bevTemp1" = "red", "line" = "blue", "bevTemp2" = "yellow" )

bev1 <- ggplot(beaver1, aes(x=time, y = temp)) + 
  geom_bar(stat = "identity", aes(colour = "bevTemp1"), fill = "red")+ 
  coord_cartesian(ylim = c(35, 38)) +
  geom_line(size = 1.5, aes(color = "line"))+
  scale_color_manual(values = colour, 
                     guide = guide_legend(override.aes = list(
                       border=c(NA, NA),
                       fill=c("red","blue"))))+
  theme(legend.title = element_blank(), legend.position = "none")
  


bev2<- ggplot(beaver2, aes(x=time, y = temp)) + 
  geom_bar(stat = "identity", aes(colour = "bevTemp2"), fill = "yellow")+ 
  coord_cartesian(ylim = c(35, 38)) +
  geom_line(size = 1.5, aes(color = "line"))+
  scale_color_manual(values = colour, 
                     guide = guide_legend(override.aes = list(
                       border=c(NA, NA),
                       fill=c("yellow","blue"))))+
  theme(legend.title = element_blank(), legend.position = "none")

cowplot::plot_grid(
  cowplot::plot_grid(
    bev1 ,
    bev2,
    align = 'h'),
  cowplot::get_legend(bev1 +  theme(legend.position = "bottom")),
  nrow = 2, rel_heights = c(4, 1))

Suggestions...

score 2 · Answer 1 · answered Jul 03 '20 at 15:47

You could build a dummy plot just to get the desired legend and add it to the final plot. I would also suggest using the line legend for the geom_line.

library(ggplot2)

# Create dummy data
df <- data.frame(class = c("bevTemp1","bevTemp2"),
                 x = 1:2,
                 y = 2:3)

# Create dummy plot just to extract the desired legend
p1<- ggplot(df, aes(x=x,y=y)) + 
  geom_col(aes(fill = class))+
  geom_line(aes(col = "line"), size = 1.5)+  
  scale_fill_manual(values = c("red","yellow")) +
  scale_color_manual(values = "blue")+
  theme(legend.title = element_blank())

cowplot::plot_grid(
  cowplot::plot_grid(
    bev1 ,
    bev2,
    align = 'h'),
  # Add the legend of the dummy plot 
  cowplot::get_legend(p1 +  theme(legend.position = "bottom")),
  nrow = 2, rel_heights = c(4, 1))

very creative! I definitely didn't think about making a dummy plot. Thanks :) — Jenna Hamlin, Jul 04 '20 at 19:47

score 1 · Answer 2 · answered Jul 03 '20 at 15:50

get_legend just retrieves the legend from the plot passed to this function, here bev1. So you need to also show bev2Temp in the legend of plot 1. You can add an additional id column to your dataset, make it a factor variable but in both datasets include bevTemp1/bevTemp2 as levels. In ggplot in scale_fill_manual, if you set drop = FALSE, all levels are shown:

library(datasets)
library(ggplot2)
library(cowplot)
data(beavers)

colour <- c("bevTemp1" = "red", "bevTemp2" = "yellow")
beaver1$id <- "bevTemp1"
beaver1$id <- factor(beaver1$id, levels = c("bevTemp1", "bevTemp2"))

beaver2$id <- "bevTemp2"
beaver2$id <- factor(beaver2$id, levels = c("bevTemp1", "bevTemp2"))


bev1 <- ggplot(beaver1, aes(x=time, y = temp, fill = id)) + 
  geom_bar(stat = "identity")+ 
  coord_cartesian(ylim = c(35, 38)) +
  geom_line(size = 1.5, aes(color = "line"))+
  theme(legend.title = element_blank(), legend.position = "none") +
  scale_fill_manual(values = c("bevTemp1" = "red", "bevTemp2" = "yellow"),
                    drop = FALSE) +
  scale_colour_manual(values = c("line" = "blue"))



bev2<- ggplot(beaver2, aes(x=time, y = temp, fill = id)) + 
  geom_bar(stat = "identity")+ 
  coord_cartesian(ylim = c(35, 38)) +
  geom_line(size = 1.5, aes(color = "line"))+
  theme(legend.title = element_blank(), legend.position = "none") +
  scale_fill_manual(values = c("bevTemp1" = "red", "bevTemp2" = "yellow"),
                    drop = FALSE) +
  scale_colour_manual(values = c("line" = "blue"))

cowplot::plot_grid(
  cowplot::plot_grid(
    bev1 ,
    bev2,
    align = 'h'),
  cowplot::get_legend(bev1 +  theme(legend.position = "bottom")),
  nrow = 2, rel_heights = c(4, 1))

yea, that makes sense, its kinda like facet wrapping by giving specific levels. Thank you :) — Jenna Hamlin, Jul 04 '20 at 19:48

make a common legend and include unique legend attributes

2 Answers2