Compare grouped boxplot with Wilcoxon test

Question

I want to add the wilcoxon test pvalue comparing the g1 vs g2 on Species subgroups to the plot below my question.

In SO there were some similar questions like this one https://stackoverflow.com/a/66627716/13046806 but the test doesn't compare 2 groups in multiples subgroups

library(dplyr)
library(ggplot2)
set.seed(123)
iris$group=sample(c("g1","g2"),nrow(iris),replace = TRUE,)
ggplot(iris,aes(x=Species ,y=Sepal.Length, fill=group) )+
  geom_boxplot()

the wilcoxon test should be something like this but ideally implemented in the ggplot pipe

L=iris %>%
  group_split(Species)
 for (i in 1:length(L)) {
   df=L[[i]]
   sub_g=unique(df$Species)
   p=df%>%
   summarise(pvalue = wilcox.test( Sepal.Length~ group,exact = FALSE)$p.value)
   print(paste0(sub_g,": ",p))
 }
#> [1] "setosa: 0.976165659102827"
#> [1] "versicolor: 0.7475079522341"
#> [1] "virginica: 0.521330527688775"

^{Created on 2023-04-24 with reprex v2.0.2}

Answer 1

You could create a new data frame with the p-values, and use it in a new geom_text layer:

irisp <- iris %>% group_by(Species) %>%
  summarise(wilcox.p = wilcox.test(Sepal.Length ~ group, exact = FALSE)$p.value)

ggplot(iris) +
  geom_boxplot(aes(x = Species, y = Sepal.Length, fill = group)) +
  geom_text(aes(x = Species,y = 3.8,label = paste0("p = ",round(wilcox.p,3))), irisp)

Answer 2

This is one way perhaps that you need

 ggplot(iris, aes(x = group, y = Sepal.Length, fill = group)) +
  geom_boxplot() +
  facet_wrap(~Species, scales = "free") +
  stat_compare_means(aes(label = paste0("p = ", ..p.format..)),
                     method = "wilcox.test", label.y = 8)