Search code examples
rggplot2geom-point

Changing specific value color with ggplot

I am using geom_point() with ggplot to essentially create the equivalent of a -coefplot- with coefficients of 5 different models on one plot.

Is it possible for me to change the color of just one point (and corresponding CI bars)? What about the color/font/etc. of an individual axis tick label? For example, if I only want to make the font bold for the label for Model 3 and make Model 3's coefficient point (+ CI bars) red (or any other color to distinguish it). I've tried scale_fill_manual but that didn't seem to do the trick.

My current code (with scale_x_discrete, the names of the models are just Model 1, 2, 3, etc. and I'm putting them in order there):

ggplot(d, aes(x = var, y = coef)) + 
  geom_point() + 
  geom_pointrange(aes(ymin = cilow, ymax = ciupper)) + 
  scale_x_discrete(limits = model_order, labels = model_label) + 
  theme_minimal()
Solution

  • You can use a condition to color only the point and CI that you want. ifelse(test to select your group, color name if yes, color name if not).

    library(tidyverse)

df=iris %>% 
  group_by(Species) %>% 
    summarise(min=min(Petal.Length),
              max=max(Sepal.Length),
              mean=(min+max)/2)  
ggplot(df,aes(Species, mean,color=Species)) +
  geom_point() +
  geom_pointrange(aes(ymin = min, ymax = max))+
  scale_color_manual(values = ifelse(df$Species=="setosa","red","black"))

    To change your axis labels, you can use ggtext as indicated in this post

    library(ggtext)
library(glue)

highlight = function(x, pat, color="black", family="") {
  ifelse(grepl(pat, x), glue("<b style='font-family:{family}; color:{color}'>{x}</b>"), x)
}
ggplot(df,aes(Species, mean,color=Species)) +
  geom_point() +
  geom_pointrange(aes(ymin = min, ymax = max))+
  scale_x_discrete(labels=function(x) highlight(x, "setosa", "purple")) +
  scale_color_manual(values = ifelse(df$Species=="setosa","red","black"))
  theme_bw()+
  theme(axis.text.x=element_markdown(size=15))

    It didn't work with theme_minimal, so I used a different one.

    Plot