Connect medians within nested groups in ggplot?-CodePudding

I am trying to plot these different groups and connect the medians of every series (A,B,C) within each group with a line.

    n <- 60
    data <- data.frame(series=rep(LETTERS[1:3], n/3), 
                   sex=rep(c("F","M"),each=30), 
                   setting=rep(c("wild","rural"),n/2),
                   fit=rnorm(n))

    ggplot(data,aes(x=sex, y=fit, fill=series))  
    geom_boxplot(width=.3,aes( alpha=.5,color=sex),
              lwd=0.8, position = position_dodge(width = 0.6))   
    facet_grid(~setting)  
    stat_summary(fun.y=median, geom="point", shape=23, size=2,
               position=position_dodge(width = 0.6))  
    geom_text(aes(y=-2.5, label=series),  position=position_dodge(width=0.6))  

    geom_point(shape=20,alpha=0.2,position=position_jitterdodge(dodge.width = 0.6,jitter.width = 0.25)) 
    theme_blank()

I have been playing around with geom_line and stats_summary, but not super succesfully, as the grouping always seems wrong. Here is an example of what I tried with stats_summary and what looks like the 4 lines that I actually want (connecting A to C in each group), but is not overlayed with the boxplots for some reason...

    stat_summary(fun.y=median, colour="red", geom="line",aes(x=series, group=interaction(setting,sex)),
             position=position_dodge(width = 0.6))

I have also found a very similar answered question here, but I could not get this soulution working for me. I get an error message when running the second part (Error in geom_line():Problem while computing aesthetics.ℹ Error occurred in the 4th layer. Caused by error in FUN():! object 'series' not found)

At this point I'd also be happy to accept any type of workaround, but couldn't come up with it myself. Thanks!

CodePudding user response：

Basically this is an adaption of the approach in the answer you referenced to your case which makes use of "manual dodging", which means to compute the x (and of course the y). positions for your median lines manually. To this end we first have to convert the sex column which is mapped on x to a numeric. Second, we have to shift the positions depending series (and of course taking account of the width by which you dodge the box plots and the points), i.e. for group A we shift the position to the left and for group C to the right.

set.seed(123)

library(ggplot2)
library(dplyr)

data_line <- data %>%
  group_by(setting, sex, series) %>%
  summarise(fit = median(fit)) |>
  ungroup() |> 
  # Manual dodging 
  mutate(x_num = as.numeric(factor(sex))   seq(-1, 1, length.out = 3) * .6 / 3)
#> `summarise()` has grouped output by 'setting', 'sex'. You can override using
#> the `.groups` argument.

ggplot(data, aes(x = sex, y = fit, fill = series))  
  geom_boxplot(
    width = .3, aes(alpha = .5, color = sex),
    lwd = 0.8, position = position_dodge(width = 0.6)
  )  
  facet_grid(~setting)  
  stat_summary(
    fun = median, geom = "point", shape = 23, size = 2,
    position = position_dodge(width = 0.6)
  )  
  geom_line(data = data_line,
    aes(x = x_num, group = sex),
    colour = "red"
  ) 
  geom_text(aes(y = -2.5, label = series), position = position_dodge(width = 0.6))  
  geom_point(shape = 20, alpha = 0.2, position = position_jitterdodge(dodge.width = 0.6, jitter.width = 0.25))
#> Warning: `position_dodge()` requires non-overlapping x intervals
#> `position_dodge()` requires non-overlapping x intervals

CodePudding user response：

This is a similar approach to Stefan's, but with the calculations done on-the-fly within ggplot:

library(ggplot2)

n <- 60
data <- data.frame(series=rep(LETTERS[1:3], n/3), 
                   sex=rep(c("F","M"),each=30), 
                   setting=rep(c("wild","rural"),n/2),
                   fit=rnorm(n))

ggplot(data, aes(x = as.numeric(interaction(series, sex))  
                   1.5 * ((as.numeric(interaction(series, sex)) - 1) %/% 3), 
                 y = fit, fill = series))  
  geom_boxplot(width = 0.5, aes(color = sex), alpha = 0.5)   
  geom_point(fun = median, stat = "summary", shape = 23, size = 2)  
  geom_text(aes(y = -2.5, label = series))  
  geom_point(shape = 20, alpha = 0.2, position = position_jitter(0.25))  
  geom_line(stat = "summary", fun = median, aes(group = sex))  
  facet_grid(~setting)  
  scale_x_continuous(NULL, breaks = c(2, 6.5), labels = c("F", "M"),
                     limits = c(0, 8.5))  
  ggnetwork::theme_blank()  
  theme(axis.text.x.bottom = element_text(), strip.background = element_blank())

^{Created on 2022-12-14 with reprex v2.0.2}