how to connect the means with a line within a single category in ggplot

Question

Here is a dummy code :

library(ggplot2)
library(dplyr)

diamonds |> dplyr::filter(color %in% c("D","E", "F"), cut %in% c("Ideal","Fair"), clarity %in% c("SI2","VS2","IF")) |> ggplot(aes(x = clarity, y =carat,  color=color, shape=cut)) +
stat_summary(fun.data= mean_cl_boot, geom="errorbar", width=0.05, position=position_dodge(0.7)) +
stat_summary(fun=mean, geom="point", size=2, position= position_dodge(0.7))

I would like to connect the means with a line within each clarity category ( ie connect circle to the triangle: shown in red colour on the picture as an example):

If I use geom_stat or geom_line: it gives an error that geom_path: Each group consists of only one observation. Do you need to adjust the group aesthetic? which makes sense since both of them are within a single clarity group. I tried to use group=interaction() but it did not work either, I only were able to do it for points within different clarity groups

This code is not reproducible. Would you care correcting it, please? Also, although not an answer to your question, I think your visualisation might become quite overwhelming. A good way to show paired data (your connected pairs) would be to use the two dimensions of your coordinate system. here carat, fair on the x and Ideal on the y axis. You could then show actually all values rather than just mean and error bars, maybe, if too crowded, 2d density estimates etc. — tjebo, May 30 '22 at 22:47
tjebo: you just have a conflict in filter function: `dplyr::filter` will fix the issue — yuliaUU, May 31 '22 at 15:04

tjebo · Accepted Answer · 2022-05-31T15:36:14.907

I think best to use a manual dodge

library(ggplot2)
library(dplyr)

df <- diamonds %>% dplyr::filter(color %in% c("D","E", "F"), cut %in% c("Ideal","Fair"), clarity %in% c("SI2","VS2","IF")) 

## make a names vector for your manual dodge 
## this of course needs adjustment depending on your actual data. can be automated
dodge_vec <- seq(-.25, .25, length = 6)
names(dodge_vec) <- unique(with(df, paste(cut, color, sep = "_")))

## some data alterations - assign dodge by subsetting with named vector
df <- df %>%
  mutate(cut_col = dodge_vec[paste(cut, color, sep = "_")]) 
## summarise for your lines 
df_line <- 
  df %>%
  group_by(clarity, cut, color, cut_col) %>%
  summarise(mean_carat = mean(carat))
#> `summarise()` has grouped output by 'clarity', 'cut', 'color'. You can override
#> using the `.groups` argument.

## need to pass your original x as an integer and add your new doding column
ggplot(df, aes(x = as.integer(factor(clarity)) + cut_col, y =carat, color=color, shape=cut)) +
stat_summary(fun.data= mean_cl_boot, geom="errorbar", width=0.05) +
  stat_summary(fun=mean, geom="point", size=2) +
  ## add lines with your new data, using an interaction variable
  geom_line(data = df_line, aes(y = mean_carat, group = interaction( as.integer(clarity), color))) +
  scale_x_continuous(breaks = 1:3, labels = unique(df$clarity))
#> Warning: Using shapes for an ordinal variable is not advised

Your question suggests that you're dealing with paired data, therefore my suggestion in the comment. I wanted to give an example, but the diamond data set doesn't have paired data, thus it would be a bit difficult to fake that.

^{Created on 2022-05-31 by the reprex package (v2.0.1)}

@PesKchan I generally don’t really recommend that visualisation. that was more to show technical feasibility. paired data is much more convincingly visualised with scatter plots. see for example my suggestions in https://stackoverflow.com/questions/70397418/create-a-split-violin-plot-with-paired-points-and-proper-orientation or https://stackoverflow.com/a/70323223/7941188 — tjebo, Jun 07 '22 at 08:00
first one is what i would incorporate since i have data designed as pre and post treatment for patient groups — PesKchan, Jun 07 '22 at 08:15

how to connect the means with a line within a single category in ggplot

1 Answers1