Add custom annotation to p-value label to ggpubr stat_compare_means()

Question

I try to make boxplot with displaying p-value

my_comparisons <- list( c("0.5", "1"), c("1", "2"), c("0.5", "2") )
ggboxplot(ToothGrowth, x = "dose", y = "len")+ 
  stat_compare_means(comparisons = my_comparisons, method = "wilcox.test")

the result is

But how can I add additional text to calculated p-values? I want to add "p = " as shown below

How can I make it?

upd. Variants below not working

my_comparisons <- list( c("0.5", "1"), c("1", "2"), c("0.5", "2") )

ggboxplot(ToothGrowth, x = "dose", y = "len")+ 
  stat_compare_means(comparisons = my_comparisons, method = "wilcox.test", aes(label=paste("p=",scientific(as.numeric(..p.format..)))))

ggboxplot(ToothGrowth, x = "dose", y = "len")+ 
  stat_compare_means(comparisons = my_comparisons, method = "wilcox.test", aes(label = paste("p =", ..p.format..)))

ggboxplot(ToothGrowth, x = "dose", y = "len")+ 
  stat_compare_means(comparisons = my_comparisons, method = "wilcox.test", aes(label = paste0("p =", ..p.format..)))

Answer 1

I believe this could be potentially a bug in ggpubr. The documentation provides an example how to add custom text to your p-value. This starts to fail when passing several comparisons to the labelling function. I think the bug might be that ggplot doesn't know which expression to parse to a label (see r-lang warning below). There are also several warnings regarding the calculation of the p-value that might be relevant.

That said, I feel that most questions regarding ggpubr are concerning annotations with multiple p-values, and I do feel that you might want to reconsider (a) your statistics and (b) your visualisation.

library(ggpubr)
#> Loading required package: ggplot2
library(tidyverse)

## This works as expected. Only one test
ggboxplot(ToothGrowth, x = "supp", y = "len") + 
  stat_compare_means(aes(label = paste0("p = ", ..p.format..)))

See plenty of warnings - starts to fail.

my_comparisons <- list( c("0.5", "1"), c("1", "2"), c("0.5", "2") )
ggboxplot(ToothGrowth, x = "dose", y = "len") +
  stat_compare_means(comparisons = my_comparisons, 
                     aes(label = paste0("p = ", ..p.format..)))
#> Warning: Using `as.character()` on a quosure is deprecated as of rlang 0.3.0.
#> Please use `as_label()` or `as_name()` instead.
#> This warning is displayed once per session.
#> Warning in wilcox.test.default(c(4.2, 11.5, 7.3, 5.8, 6.4, 10, 11.2, 11.2, :
#> cannot compute exact p-value with ties
#> Warning in wilcox.test.default(c(4.2, 11.5, 7.3, 5.8, 6.4, 10, 11.2, 11.2, :
#> cannot compute exact p-value with ties
#> Warning in wilcox.test.default(c(16.5, 16.5, 15.2, 17.3, 22.5, 17.3, 13.6, :
#> cannot compute exact p-value with ties

By creating a plot for each compared group separately, this starts to work again.

ls_tg <- ToothGrowth %>%
  split(., .$dose)

lapply(my_comparisons, function(x) bind_rows(ls_tg[x])) %>%
  map(~ggboxplot(., x = "dose", y = "len") + 
        stat_compare_means(aes(label = paste0("p = ", ..p.format..)))) %>%
  patchwork::wrap_plots()

^{Created on 2021-12-07 by the reprex package (v2.0.1)}

Answer 2

They provide also a workaround, I changed it a bit to create your own custom formatted.p. Note that you need the package rstatix as well.

library(ggpubr)
library(ggplot2)
library(rstatix)

data("ToothGrowth") # data is part of the package ggpubr

my_comparisons = list(c("0.5", "1"), c("1", "2"), c("0.5", "2"))

stat.test <- ToothGrowth %>% 
  wilcox_test(len ~ dose, comparisons = my_comparisons) %>%
  add_xy_position(x = "dose") %>%
  mutate(myformatted.p = paste0("p = ", p))

stat.test

# # A tibble: 3 x 14
#   .y.   group1 group2    n1    n2 statistic            p       p.adj p.adj.signif y.position groups     xmin  xmax myformatted.p   
#   <chr> <chr>  <chr>  <int> <int>     <dbl>        <dbl>       <dbl> <chr>             <dbl> <list>    <dbl> <dbl> <chr>           
# 1 len   0.5    1         20    20      33.5 0.00000702   0.000014    ****               35.4 <chr [2]>     1     2 p = 0.00000702  
# 2 len   1      2         20    20      61   0.000177     0.000177    ***                37.6 <chr [2]>     2     3 p = 0.000177    
# 3 len   0.5    2         20    20       1.5 0.0000000841 0.000000252 ****               39.9 <chr [2]>     1     3 p = 0.0000000841


ggboxplot(ToothGrowth, x = "dose", y = "len", color = "dose") + 
  stat_pvalue_manual(stat.test, label = "myformatted.p")

To get more control over the format and rounding you could use this or alter to any format you wish.

mutate(myformatted.p = paste0("p = ", ifelse(p < 1e-5, format(p, scientific = T), signif(p, digits = 2))))