How to add effect sizes to ggplot bar graphs of t-tests? (e.g., Cohen's d or Hedges' g)

Question

I have a dataset like the following.

Group <- c("A", "B", "A", "A", "A", "B", "A", "A", "B", "B","A", "A", "B", "B")
Score <- c(26, 22, 15, 5, 19, 3, 4, 5, 23, 3, 5, 2, 20, 4)
Order <- c("First", "First", "First", "Second", "First", "Second", "Second", "Second", "First", "Second", "Second", "Second", "First", "Second")
Data <- data.frame(Group, Score, Order)

Data

   Group Score  Order
1      A    26  First
2      B    22  First
3      A    15  First
4      A     5 Second
5      A    19  First
6      B     3 Second
7      A     4 Second
8      A     5 Second
9      B    23  First
10     B     3 Second
11     A     5 Second
12     A     2 Second
13     B    20  First
14     B     4 Second

I need to plot the differences in scores between each group, with effect sizes for each comparison (either Cohen's d or Hedge's g would work).

The code below gets me everything I want on the plot, but the effect size.

OrderComparison <- list(c("First", "Second"))

ggplot(Data, aes(Order, Score, fill=Order))+ 
  stat_summary(geom = "bar", fun = mean, position = "dodge", color="black")+
  stat_summary(geom = "errorbar", fun.data = mean_se, position = "dodge", width=.2)+
  stat_compare_means(method = "t.test", comparisons = OrderComparison, label = "p.signif", position = "identity")+
  facet_wrap(~Group, scales="fixed", strip.position = "bottom")+
  theme_classic()+
  theme(legend.position = "none")+
  scale_fill_grey(start = .6, end = 1)

Question

What code (and/or package) do I need to automatically put an effect size on each comparison?

Any advice or guidance is appreciated.

Answer 1

library(effsize)
library(dplyr)

cohen=numeric()
for (i in LETTERS[1:2]) {
  Group=filter(Data, Group==i)
  treatment = filter(Group, Order=="First")%>%select(Score)%>%unlist()
  control = filter(Group, Order=="Second")%>%select(Score)%>%unlist()
  d = (c(treatment,control))
  f = c(rep("Treatment", length(treatment)), rep("Control", length(control)))
  c=cohen.d(treatment,control)
  cohen[i]=c$estimate
}

effsize=data.frame(sz=c(rep(paste("Cohen's d:", cohen[1]),8), rep(paste("Cohen's d:", cohen[2]),6)), Group=c(rep("A", 8), rep("B", 6)))

ggplot(Data, aes(Order, Score, fill=Order))+ 
  stat_summary(geom = "bar", fun = mean, position = "dodge", color="black")+
  stat_summary(geom = "errorbar", fun.data = mean_se, position = "dodge", width=.2)+
  stat_compare_means(method = "t.test", comparisons = OrderComparison, label = "p.signif", position = "identity")+
  facet_wrap(~Group, scales="fixed", strip.position = "bottom")+
  theme_classic()+
  theme(legend.position = "none")+
  scale_fill_grey(start = .6, end = 1)+
  geom_text(aes(x=1.5,y=25, label=sz), data=effsize)

It manually calculates the Cohen's d for the two groups and puts it in the plot.