Adding p-values to compare groups means at different at times in gganimate gif with boxplots/violins

Question

My current code for printing violins using gganimate looks like this

  library(ggplot2); library(gganimate); library(ggpubr)
  ggplot(dat2, aes(x=diet, y=bicep, fill=diet)) + 
  geom_violin() +
  scale_fill_manual(values=c("#00AFBB", "#FC4E07")) +
  stat_compare_means(aes(label = ..p.format..), paired = FALSE, label.x.npc = 0.5) +
  labs(title = 'Week: {frame_time}') +
  transition_time(time) +
  ease_aes('linear')

Here the p values are printed but they are just overall p values. I would like the p-value to change over time (0, 6 and 12 weeks). In my study the each outcome measurement (bicep) is taken at three different times (0, 6 and 12 weeks or time 1, time 2, time 3), It would be neat if I could show changing p-values at time 0, 6, 12. Here I would use a unpaired t test to compare group means across diet/treatment.

Alternatively, show p-value (paired t test) just at the end where bicep at time '3' is compared to bicep at time '1' for both diets.

How would I go about doing this? Thanks for reading this.

Data Structure

 structure(list(code = c(1L, 1L, 1L, 2L, 2L, 2L, 3L, 3L, 3L, 4L, 
4L, 4L, 5L, 5L, 5L, 6L, 6L, 6L, 7L, 7L, 7L, 8L, 8L, 8L), diet = c("a", 
"a", "a", "b", "b", "b", "a", "a", "a", "b", "b", "b", "a", "a", 
"a", "b", "b", "b", "a", "a", "a", "b", "b", "b"), time = c(1L, 
2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 3L, 1L, 2L, 
3L, 1L, 2L, 3L, 1L, 2L, 3L), bicep = c(8L, 7L, 7L, 9L, 9L, 9L, 
11L, 10L, 9L, 11L, 11L, 12L, 12L, 11L, 10L, 9L, 9L, 9L, 12L, 
10L, 8L, 12L, 12L, 12L)), class = "data.frame", row.names = c(NA, 
-24L))

Reproducible gganimate code

    ggplot(example3, aes(x=diet, y=bicep, fill=diet)) + 
  geom_violin() +
  scale_fill_manual(values=c("#00AFBB", "#FC4E07")) +
  stat_compare_means(aes(label = ..p.format..), paired = FALSE, label.x.npc = 0.5) +
  labs(title = 'Week: {frame_time}') +
  transition_time(time) +
  ease_aes('linear')

The data example you provided doesn't seem to have the same fields as what you use in the code. I'm interested in trying to help but this code doesn't run on its own. — Jon Spring, Nov 18 '18 at 23:46
@JonSpring thanks for checking, I just looked over the example code and actual code a few times. I can't seem the notice the differences. Could you please point me towards what is different? I will correct it. — DiscoR, Nov 19 '18 at 00:50
@JonSpring I have changed the reproducible data and example code and tested it. It works at my end. Thank you. — DiscoR, Nov 19 '18 at 01:00

Roman · Accepted Answer · 2018-11-19T14:13:18.867

0

You can try to calculate the p.values in advance.

library(gganimate)
library(tidyverse)
example3 %>%
  group_by(time) %>% 
  mutate(p=wilcox.test(bicep~diet, exact =F)$p.value,
         max=max(bicep, na.rm = T)) %>% 
  ggplot() + 
  geom_violin(aes(x=diet, y=bicep, fill=diet)) +
  geom_text(data = . %>% distinct(p, max, time), 
            aes(x=1.5, y = max+.5, label=as.character(round(p,2))),
            size=12) +
  transition_time(time) +
  ease_aes('linear')

You need to install ggpubr_0.2

edited Nov 19 '18 at 14:13

answered Nov 19 '18 at 10:53

Roman

17,008
3
36
49

This is great, thanks! Just to check this is comparing the two group means Diet a vs Diet b right? What would I have to do compare Diet 'a' bicep mean at time 3 to Diet 'a' bicep reading at time 1? And the same for Diet b? – DiscoR Nov 19 '18 at 17:19
just do `pairwise.wilcox.test(example3$bicep, interaction(example3$diet, example3$time), p.adjust.method = "none")` – Roman Nov 20 '18 at 08:58

Adding p-values to compare groups means at different at times in gganimate gif with boxplots/violins

1 Answers1