Search code examples
rggplot2violin-plot

Half violin plot with different factors in R

I have data of a pre and a posttest in two groups: control and treatment. I computed the score and now want to illustrate the differences between groups and tests. As a basis I have a data frame in R containing for each student a row with the results of the pre as well as the posttest:

student_id  group       test_id  Score
145         Treatment   pre      0.12
145         Treatment   post     0.78
109         Control     pre      0.45
109         Control     post     0.99

I have written the following code

s2$test_id <- as.factor(s2$test_id)
s2$group <- as.factor(s2$group)

p <- ggplot(s2, aes(x = test_id, y=score, fill = group)) + geom_violinhalf(trim=FALSE, fill="gray")+
  labs(title="Half Violin",x="Test", y = "Score")+
  geom_boxplot(width=0.1)+
  theme_classic() + scale_x_discrete(limits = rev)
p

producing

half_violin

Is it possible, that the grey violin can be changed into the distribution of the treatment group (blue) and that I can add a second half violin shape on the left of the pre and posttest depicting the control distribution (red)?

Solution

  • You can use the same grouping variable that you use for the boxplots inside geom_violinhalf, then use the flip argument to flip the correct ones around:

    library(ggplot2)
library(see)

ggplot(s2, aes(test_id, Score, fill = group)) +
  geom_boxplot(width = 0.1, position = position_dodge(0.2)) +
  geom_violinhalf(aes(group = interaction(test_id, group)), fill = 'gray',
                  trim = FALSE, flip = c(1, 2)) +
  theme_classic(16)

    enter image description here

    Data used based on structure given in snippet in question

    set.seed(1)

s2 <- data.frame(student_id = rep(sample(500, 50), 2),
                 group = rep(sample(c('Treatment', 'Control'), 50, TRUE), 
                             each = 2),
                 test_id = rep(c('pre', 'post'), 50),
                 Score = runif(100))