Scale geom_point size to increase size based on distance from zero


Hi folks!

I'd like to plot some measures that have been standardized to z-scores. I want the size of the point in geom_point() to increase from 0 to 3, and also to increase from 0 to -3. I also want the colour to change from red, to blue. The trick is to get both to work together.

Here is an example that's as close as I can get to what I'd like, note that the size of the point increases from -2, whereas I want the size of the point to increase as the z_score moves away from zero.


year <- rep(c(2015:2018), each = 3)
parameters <- rep(c("length", "weight", "condition"), 4)
z_score <- runif(12, min = -3, max = 3)
df <- tibble(year, parameters, z_score)

cols <- c("#d73027",

ggplot(df, aes(year, parameters, colour = z_score, size = z_score)) +
  geom_point() +
  scale_colour_gradientn(colours = cols) +
  theme(legend.position="bottom") +
  scale_size(range = c(1,15)) +
  guides(color= guide_legend(), size=guide_legend())


One trick I tried was to use the absolute value of z_score which scaled the points correctly but messed up the legend.

Here's what I'd like the legend and points size to be scaled to, though I'd like the colour to be a gradient as in my example. Any insight would be greatly appreciated!



You can override the legend guides to make it work. The idea is that you need to specify the number of legend's size.


year <- rep(c(2015:2018), each = 3)
parameters <- rep(c("length", "weight", "condition"), 4)
z_score <- runif(12, min = -5, max = 5)
df <- tibble(year, parameters, z_score)

# The size of the legends
legend_size <- c(10,8,5,1,5,8,10)

cols <- c("#d73027",

ggplot(df, aes(year, parameters)) +
  geom_point(aes(colour = z_score, size = z_score)) +
  scale_color_continuous(low = "blue", high = "red", breaks = c(-3:3) ) +
  scale_size(breaks = c(-3:3), range = c(1,15)) +
    color= guide_legend(), 
    size=guide_legend(override.aes = list(size = legend_size))

Created on 2018-12-19 by the reprex package (v0.2.1)

I wonder why would you want to use 2 aesthetics (size and color) for 1 variable? Is there any extra information added compare to only 1 aesthetic?.

This answer was based on a StackOverflow answer


This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.