geom_pwc label = "p.adj" breaks with faceting

[marklyng]

With non-faceted data, geom_pwc reads label = "p.adj" and appends the adjusted p-value to the plot. When faceting by a variable, label = "p.adj" is ignored and the unadjusted p-value is used instead.

No facet

x <- tibble(value = c(seq(10), seq(10, 100, 10), seq(100, 1000, 100), seq(1000, 10000, 1000)),
            variable = c(rep("a", 10), rep("b", 10), rep("c", 10), rep("d", 10)))


compare_means(value ~ variable, data = x)
# A tibble: 6 × 8
  .y.   group1 group2         p    p.adj p.format p.signif method  
  <chr> <chr>  <chr>      <dbl>    <dbl> <chr>    <chr>    <chr>   
1 value a      b      0.000211  0.00063  0.00021  ***      Wilcoxon
2 value a      c      0.0000108 0.000065 1.1e-05  ****     Wilcoxon
3 value a      d      0.0000108 0.000065 1.1e-05  ****     Wilcoxon
4 value b      c      0.000211  0.00063  0.00021  ***      Wilcoxon
5 value b      d      0.0000108 0.000065 1.1e-05  ****     Wilcoxon
6 value c      d      0.000211  0.00063  0.00021  ***      Wilcoxon

x %>% 
  ggplot(aes(x = variable, y = value)) +
  geom_boxplot() +
  geom_pwc(label = "p.adj")

Facet

y <- tibble(value = c(seq(10), seq(10, 100, 10), seq(100, 1000, 100), seq(1000, 10000, 1000)),
            variable = c(rep("a", 10), rep("b", 10), rep("a", 10), rep("b", 10)),
            facet_variable = c(rep("test1", 20), rep("test2", 20)))

compare_means(value ~ variable, data = y, group.by = "facet_variable")
# A tibble: 2 × 9
  facet_variable .y.   group1 group2        p   p.adj p.format p.signif method  
  <chr>          <chr> <chr>  <chr>     <dbl>   <dbl> <chr>    <chr>    <chr>   
1 test1          value a      b      0.000211 0.00042 0.00021  ***      Wilcoxon
2 test2          value a      b      0.000211 0.00042 0.00021  ***      Wilcoxon

y %>% 
  ggplot(aes(x = variable, y = value)) +
  geom_boxplot() +
  geom_pwc(label = "p.adj") +
  facet_grid(. ~ facet_variable)

Changing the label to any of the other "p.adj.*" types does not fix the issue.

[kassambara]

When using the ggplot facet functions, the p-values are computed and adjusted by panel, without taking into account the other panels. This is by design in ggplot2.

This means that, when there is only one computed p-value by panel, then using label = "p" or label = "p.adj" will give the same results using geom_pwc(). P-value computation and adjustment in a given facet panel is done independently to the other panels.

One might want to adjust the p-values of all the facet panels together. There are two solutions for that:

1. Using ggadjust_pvalue(p) after creating the plot p. This function is available in the current github dev version, which can be installed using devtools::install_github("kassambara/ggpubr")
2. or adding the adjusted p-value manually using stat_pvalue_manual(). Read more at:
   • How to Add P-values to GGPLOT Facets
   • Add P-values to GGPLOT Facets with Different Scales

The following example shows how to use ggadjust_pvalue(p):

suppressPackageStartupMessages(library(ggpubr))

y <- tibble::tibble(value = c(seq(10), seq(10, 100, 10), seq(100, 1000, 100), seq(1000, 10000, 1000)),
            variable = c(rep("a", 10), rep("b", 10), rep("a", 10), rep("b", 10)),
            facet_variable = c(rep("test1", 20), rep("test2", 20)))
p <- y %>% 
  ggplot(aes(x = variable, y = value)) +
  geom_boxplot() +
  geom_pwc(label = "p.adj") +
  facet_grid(. ~ facet_variable)
# Adjust  all panels p-values together
ggadjust_pvalue(p, label = "p.adj")