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}

you want comparison between setosa(g1,g2)? if i see the code you are now specifying any test to compare here — PesKchan, Apr 24 '23 at 08:43
have a look at the code i guess that is what you are looking for — PesKchan, Apr 24 '23 at 09:15

score 1 · Answer 1 · answered Apr 24 '23 at 09:14

1

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)

answered Apr 24 '23 at 09:14

PesKchan

868
6
14

that is not bad but coud the plot be unwrapped ? – Wael Apr 24 '23 at 09:19
1

that is what Im trying ,as of now my brain couldn't figure it out how to keep all of them together – PesKchan Apr 24 '23 at 09:20
@Wael https://stackoverflow.com/questions/46446392/r-ggplot2-boxplots-with-significance-level-more-than-2-groups-kruskal-test-an?rq=1 this is one i found but again its using facet – PesKchan Apr 24 '23 at 09:32

score 1 · Accepted Answer · answered Apr 24 '23 at 11:32

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)

Compare grouped boxplot with Wilcoxon test

2 Answers2