Legend for a plot with sec.axis (geom bar + geom line)

Question

I have a ggplot with two y-axes by the sec.axis command, and I've been struggling with handling legends in these situations.

The code:

library(ggplot2)
library(ggrepel)

df <- data.frame(day = as.character(seq(from = 1, to = 100, by = 1)),
             total = rbinom(n=100,30,0.5),
             prop = runif(100))

df <- df %>% arrange(df, by = day)
df$`percentage` <- label_percent(accuracy = 0.01)(df$prop)


ggplot(data = df,
       aes(x = day, y = total)) +
  geom_bar(aes(x = day, y = total), stat = "identity", fill = "lightgreen", width = 0.35) +
  geom_line(data = df,
            aes(x = day, y = (prop)*15, group = 1, color = prop),
            color = "red", size = 1,inherit.aes = TRUE) +
  scale_y_continuous(
    labels = function(x) format(x, scientific = FALSE),
    #breaks = seq(from = 0, to = 10000000,by = 100000),
    sec.axis = sec_axis(trans = ~./15,
                        name = "Secondary axis",
                        breaks = seq(from = 0, to = 10, by = 0.1),
                        scales::percent))+
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5))+
  geom_label_repel(data=df[nrow(df),],
                   aes(x = day,
                       y = prop*15,
                       label = round(prop*100,2)),
                   color = 'red',
                   nudge_x = 2,
                   segment.alpha = 0.5) +
  scale_x_discrete(expand = expansion(add = c(0, 7)))

And I wanted to simply have the legend, instead of having the axis description, like this:

I know it seems reasonably easy to obtain, but given the fact that I don’t have any groups, I either: can't plot any legend what so ever; or I get plotted two squares (when I wanted the legend to consist of a line, with the geom_line color and a square with the geom_bar color).

With the code @divibisan provided, I get the following output:

Final update:

I finally found the solution. I still have no idea how I got a different output from what @divibisan posted, but here goes what worked for me:

library(dplyr)
library(ggplot2)
library(ggrepel)

df <- data.frame(day = as.character(seq(from = 1, to = 100, by = 1)),
                 total = rbinom(n=100,30,0.5), 
                 prop = runif(100))

df <- df %>% arrange(df, by = day)
df$`percentage` <- scales::label_percent(accuracy = 0.01)(df$prop)


ggplot(data = df, 
       aes(x = day, y = total)) +
  geom_bar(aes(x = day, y = total, fill = "Total"), stat = "identity", width = 0.35) + 
  geom_line(data = df, 
            aes(x = day, y = (prop)*15, group = 1, color = 'Percentage'), size = 1,inherit.aes = TRUE) +
  scale_y_continuous( 
    labels = function(x) format(x, scientific = FALSE),
    #breaks = seq(from = 0, to = 10000000,by = 100000),
    sec.axis = sec_axis(trans = ~./15, 
                        name = "Secondary axis",
                        breaks = seq(from = 0, to = 10, by = 0.1),
                        scales::percent))+
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5))+
  geom_label_repel(data=df[nrow(df),], 
                   aes(x = day, 
                       y = prop*15,
                       label = round(prop*100,2)),
                   color = 'red',
                   nudge_x = 2,
                   segment.alpha = 0.5) +
  scale_x_discrete(expand = expansion(add = c(0, 7))) +
  scale_fill_manual(values=c('Total' = 'lightgreen'), drop=TRUE, name='') +
  scale_color_manual(values=c('Percentage' = "red"), drop=TRUE, name='') + 
  theme(legend.title = element_blank())

Answer 1

You just need to set the color/fill with a value in the aes, then use a scale function to set the color and create a legend. Here, we move the color= and fill= values from the bar and line into the aes. Then we add scale_fill/color_manual functions that set the color based on those names:

library(dplyr)
library(ggplot2)
library(ggrepel)

df <- data.frame(day = as.character(seq(from = 1, to = 100, by = 1)),
                 total = rbinom(n=100,30,0.5), 
                 prop = runif(100))

df <- df %>% arrange(df, by = day)
df$`percentage` <- scales::label_percent(accuracy = 0.01)(df$prop)


ggplot(data = df, 
       aes(x = day, y = total)) +
    geom_bar(aes(x = day, y = total, fill = "Total"), stat = "identity", width = 0.35) + 
    geom_line(data = df, 
              aes(x = day, y = (prop)*15, group = 1, color = 'Percentage'), size = 1,inherit.aes = TRUE) +
    scale_y_continuous( 
        labels = function(x) format(x, scientific = FALSE),
        #breaks = seq(from = 0, to = 10000000,by = 100000),
        sec.axis = sec_axis(trans = ~./15, 
                            name = "Secondary axis",
                            breaks = seq(from = 0, to = 10, by = 0.1),
                            scales::percent))+
    theme(axis.text.x = element_text(angle = 90, vjust = 0.5))+
    geom_label_repel(data=df[nrow(df),], 
                     aes(x = day, 
                         y = prop*15,
                         label = round(prop*100,2)),
                     color = 'red',
                     nudge_x = 2,
                     segment.alpha = 0.5) +
    scale_x_discrete(expand = expansion(add = c(0, 7))) +
    scale_fill_manual(values=c('Total' = 'lightgreen', 'Percentage'='red'), drop=TRUE, name='') +
    scale_color_manual(values=c('Total' = 'lightgreen', 'Percentage'='red'), drop=TRUE, name='')

If, for some reason, the drop argument isn't working and both colors show up in both scales, there's really no reason to include them in the scale if they're not expected to be there. Just only include the colors in the scale that are desired:

    scale_fill_manual(values=c('Total' = 'lightgreen'), drop=TRUE, name='') +
    scale_color_manual(values=c('Percentage'='red'), drop=TRUE, name='')