Pairing Scatter Points with Lines Across Fill Condition of Violin Plot

Question

Here is a data.frame with three subjects in a 2 by 2 design:

x<-data.frame("sub" = rep(c("sub1","sub2","sub3"),times=4),
          "cond1" = rep(c("A","B"),times=c(6,6)),
          "cond2" = rep(c("C","C","C","D","D","D"),times=2),
          "score" = c(6,5,4, 5,4,3, 4,3,2, 3,2,1))

Here is the split violin plot (for function, https://gist.github.com/Karel-Kroeze/746685f5613e01ba820a31e57f87ec87):

dodge <- position_dodge(width=.5)
ggplot(x, aes(x=cond1, y=score, fill=cond2))+
  geom_split_violin(trim = F)+
  geom_point(shape=16,position=dodge)

What I would like to do is connect the individual subject scatter points with a line across the fill condition (e.g., sub1 in cond1 A and cond2 C will connect to sub1 in cond1 A and cond2 D). Does anyone know if this is possible?

Answer 1

Here's an approach. First, I spread the table so the beginning and end of geom_segment can be based on different columns. Then I convert the cond1 values to numeric (which is what ggplot does under the hood) and feed those into the x fields. The x dodge offset of 0.12 was manual, but there's probably a clever way (beyond my current understanding) to use the ggplot settings to determine that automatically.

library(tidyverse)
dodge <- position_dodge(width=.5)
ggplot(x, aes(x=cond1, y=score, fill=cond2))+
  geom_split_violin(trim = F)+
  geom_point(shape=16,position=dodge) +
  geom_segment(data = x %>% spread(cond2, score),
               aes(x    = as.numeric(cond1) - 0.12, 
                   xend = as.numeric(cond1) + 0.12,
                   y = C, yend = D), inherit.aes = FALSE, color = "black")