ggpubr stat_compare_means: Show significance levels with two grouping variables

Question

I am trying to visualize significance levels (asterisks) with ggpubr 's stat_compare_means() . I encountered the following issue: As opposed to compare_means() , you cannot add a grouping variable to the comparison. Example:

ggbarplot(ToothGrowth, x = "dose", y = "len", add = "mean_se", color = "supp",
fill = "supp",position = position_dodge(0.8),add.params = list(group = "supp"))+
stat_compare_means(ref.group = "0.5", group.by = "supp",label = "p.signif")

As you can see, the significance levels are not displayed above all the bars, but only above the different doses, because ggpubr doesn't differentiate between different supps.

Is there any way to compare to these (sub)sets as well?

Thank you

Answer 1

You can try

library(tidyverse)
library(ggsignif)
ToothGrowth %>% 
     mutate(gr=interaction(supp, dose, sep = " ")) %>% 
    {ggplot(data=.,aes(x = gr,  y = len, fill = supp)) +
            stat_summary(fun.y = mean, geom = "bar") +
            stat_summary(aes(col = supp), fun.data = "mean_se", geom = "errorbar", width=0.6)+
            ggsignif::geom_signif(comparisons = combn(sort(unique(as.character(.$gr))),2, simplify = F),
                                  step_increase = 0.08,test = "wilcox.test", test.args = list(exact = FALSE))}

By adding map_signif_level = TRUE or map_signif_level = c("***"=0.001, "**"=0.01, "*"=0.05), you get