How can I combine two ggplot2 custom functions makig the legend separate

June 19, 2023

I have two custom functions:

plt_regression_line <- function(color = "black", linetype = "dashed"){
  list(
    ggplot2::geom_smooth(
      method = "lm",
      se = FALSE,
      formula = "y ~ x",
      mapping = ggplot2::aes(alpha = "Regression line"),
      linetype = linetype,
      color = color),
    ggplot2::labs(alpha = ""),
    ggplot2::scale_alpha_manual(values = c(1,1))
  )
}

and

plt_identity_line <- function(color = "black", linetype = "dashed"){
  list(
    ggplot2::geom_abline(
      mapping = ggplot2::aes(slope = 1, intercept = 0,alpha = "Identity line"),
      linetype = linetype,
      color = color),
    ggplot2::labs(alpha = ""),
    ggplot2::scale_alpha_manual(values = c(1,1))
  )
}

They work perfect individually, but when I combine them the legend does not distinguish on from another:

library(ggplot2)

  ggplot(mtcars,aes(qsec,mpg))+
  geom_point()+
  plt_regression_line(color = "red",linetype = 1)+
  plt_identity_line(color = "blue",linetype = 1)

*I use alpha, since col and fill can be used for others variables and I want to show them in the legend.

>Solution :

You are using the ‘alpha’ aesthetic twice, so ggplot is combining the legends. One potential option is to use another aesthetic, e.g. linewidth, for one of the functions:

library(tidyverse)

plt_regression_line <- function(color = "black", linetype = "dashed"){
  list(
    ggplot2::geom_smooth(
      method = "lm",
      se = FALSE,
      formula = "y ~ x",
      mapping = ggplot2::aes(alpha = "Regression line"),
      linetype = linetype,
      color = color),
    ggplot2::labs(alpha = ""),
    ggplot2::scale_alpha_manual(values = c(1,1))
  )
}

plt_identity_line <- function(color = "black", linetype = "dashed"){
  list(
    ggplot2::geom_abline(
      mapping = ggplot2::aes(slope = 1, intercept = 0,linewidth = "Identity line"),
      linetype = linetype,
      color = color),
    ggplot2::labs(linewidth = ""),
    ggplot2::scale_linewidth_manual(values = c(0.5,0.5))
  )
}

ggplot(mtcars,aes(qsec,mpg))+
  geom_point()+
  plt_regression_line(color = "red",linetype = 1) +
  plt_identity_line(color = "blue",linetype = 1)

^{Created on 2023-06-19 with reprex v2.0.2}

Another potential option is to use the ggnewscale package, e.g.

library(tidyverse)
library(ggnewscale)

plt_regression_line <- function(color = "black", linetype = "dashed"){
  list(
    ggplot2::geom_smooth(
      method = "lm",
      se = FALSE,
      formula = "y ~ x",
      mapping = ggplot2::aes(alpha = "Regression line"),
      linetype = linetype,
      color = color),
    ggplot2::labs(alpha = ""),
    ggplot2::scale_alpha_manual(values = c(1,1))
  )
}

plt_identity_line <- function(color = "black", linetype = "dashed"){
  list(
    ggplot2::geom_abline(
      mapping = ggplot2::aes(slope = 1, intercept = 0,alpha = "Identity line"),
      linetype = linetype,
      color = color),
    ggplot2::labs(alpha = ""),
    ggplot2::scale_alpha_manual(values = c(1,1))
  )
}

ggplot(mtcars,aes(qsec,mpg))+
  geom_point()+
  plt_regression_line(color = "red",linetype = 1) +
  ggnewscale::new_scale(new_aes = "alpha") +  
  plt_identity_line(color = "blue",linetype = 1)