ggplot generating two legends when only one is wanted

Question

In R I'm trying to generate a plot where I want to apply unique colors, line types, transparencies, and line thicknesses by case grouping. As currently implemented two legend plots are generated instead of one. The second legend plot is the only one that I can change the legend title. Presumably I've made a mistake any help would be greatly appreciated.

Ultimately I want to generate a single legend and have the style changes and labeling changes take effect.

library(ggplot2)
temp_df <- data.frame(year = integer(50), value = numeric(50), case = character(50))
temp_df$year <- 1:50
temp_df$value <- runif(50)
temp_df$case <- "A"
df <- temp_df
temp_df$value <- runif(50)
temp_df$case <- "B"
df <- rbind(df, temp_df)
LineTypes <- c("solid", "dotted")
colors <- c("red", "black")
linealphas <- c(1, .8)
linesizes <- c(1, 2)
Plot <- ggplot(df, aes(x = year, y = value, group = case))+
  geom_line(aes(linetype = case, color = case, size = case, alpha = case))+
  scale_linetype_manual(values = LineTypes)+
  scale_color_manual(values = colors)+
  scale_y_continuous(limits = c(0, 1), labels = scales::percent)+
  scale_alpha_manual(values = linealphas)+
  scale_size_manual(values = linesizes)+
  xlab("Year")+
  ylab("Percentage%")+
  labs(color = "Scenario")+
  theme_minimal()
Plot

You can use the `name = ` argument inside your manual scale_* functions to change the legend title, and _e.g._ `+ guides(linetype = FALSE)` to hide the legend associated with a scale_* function. But personally, I would use only colour to distinguish groups in this case. — neilfws, May 08 '18 at 23:27

score 3 · Answer 1 · answered May 09 '18 at 05:01

If you want ggplot to merge the legends then they all have to have the same title. You can specify the legend title with the name argument in the scales:

ggplot(df, aes(x = year, y = value, group = case))+
  geom_line(aes(linetype = case, color = case, size = case, alpha = case)) +
  scale_linetype_manual(values = LineTypes, name = "Scenario")+
  scale_color_manual(values = colors, name = "Scenario")+
  scale_y_continuous(limits = c(0, 1), labels = scales::percent)+
  scale_alpha_manual(values = linealphas, name = "Scenario")+
  scale_size_manual(values = linesizes, name = "Scenario")+
  xlab("Year")+
  ylab("Percentage%")+
  theme_minimal()

score 2 · Answer 2 · edited May 09 '18 at 05:05

A coworker pointed out a resolution to me the key was to remove the guides so that only one of styles that I had defined was being used for the legend.

     guides(size = FALSE)+
     guides(alpha = FALSE)+
     guides(linetype = FALSE)+

His explanation for this was that R doesn't recognize that the vector of factors defining the properties of the plot are necessarily related. As a result it will generate multiple legends when only one is desired.

    library(ggplot2)
    temp_df<-data.frame(year=integer(50),value=numeric(50),case=character(50))
    temp_df$year<-1:50
    temp_df$value<-runif(50)
    temp_df$case<-"A"
    df<-temp_df
    temp_df$value<-runif(50)
    temp_df$case<-"B"
    df<-rbind(df,temp_df)
    LineTypes<-c("solid","dotted")
    colors<-c("red","black")
    linealphas<-c(1,.8)
    linesizes<-c(1,2)
    Plot<-ggplot(df,aes(x=year,y=value,group=case))+
      geom_line(aes(linetype=case, color=case, size=case, alpha =case))+
      scale_linetype_manual(values=LineTypes)+
      scale_color_manual(values=colors)+
      scale_y_continuous(limits=c(0,1),labels = scales::percent)+
      scale_alpha_manual(values=linealphas)+
      scale_size_manual(values=linesizes)+
      xlab("Year")+
      ylab("Percentage%")+
      labs(color = "Scenario")+
      guides(size = FALSE)+
      guides(alpha = FALSE)+
      guides(linetype = FALSE)+
      theme_minimal()
    Plot

But then the legend key (`linetype`) for scenario `B` is not totally correct, is it? See [this answer](https://stackoverflow.com/a/50240153/786542) for the fix — Tung, May 08 '18 at 23:38

score 0 · Answer 3 · edited May 09 '18 at 04:57

0

Can't you just remove the line "labs(color = "Scenario")"? This is the plot that gets generated. Not sure if it's missing anything that you need.

The result seems fine to me: (seems fine to me)

edited May 09 '18 at 04:57

Claus Wilke

16,992
7
53
104

answered May 08 '18 at 23:11

Zheng Yan

9
1

Two things, I want to change the title of the legend. Yes technically I could just give the column in question a different name but I'd prefer to do it in the plotting step. Also the style changes in the plot aren't propagating to both legends. – vcelloho May 08 '18 at 23:14
Certainly, it is a work around. However, I don't know why we can not modify the legend title in this example. I tried without success – lcgodoy May 08 '18 at 23:14

ggplot generating two legends when only one is wanted

3 Answers3