r one sided geom_quasirandom-CodePudding

I am interested in plotting the distribution of some continuous variables by groups using the geom_quasirandom() function.

I would like to know how do plot only one side, for example only points on the right side rather than on both sides.

ggplot(iris,aes(Species, Sepal.Length))   geom_quasirandom()

I am interested in a plot like this. I am not interested in solutions based on geom_beeswarm because I have a lot of datapoints and geom_beeswarm just creates a crowded mess for my dataset. So any solution specific to geom_quasirandom() will be helpful. Thanks.

CodePudding user response：

Here's one approach, pulling the guts out of a beeswarm and using dplyr to adjust these manually, and adding the Species labels back manually.

ggplot_build(ggplot(iris,aes(Species, Sepal.Length))   
  geom_quasirandom()) -> a

a$data[[1]] %>%
  group_by(group) %>%
  mutate(x_adj = group   abs(x - median(x))) %>%
  ungroup() %>%
  
  ggplot(aes(x_adj, y))  
  geom_point()  
  scale_x_continuous(breaks = 1:3, labels = unique(iris$Species))

...or a hack to keep the automatic labeling by using an invisible layer:

...
ggplot(aes(x, y))  
  geom_point(alpha = 0)  
  geom_point(aes(x = x_adj))

CodePudding user response：

It's funny, the first example below is actually in the README for {ggbeeswarm} but it doesn't seem to work with my versions of R and the package.

library(tidyverse)
library(ggbeeswarm)

# doesn't look asymetric as shown in the vingette
ggplot(iris,aes(Species, Sepal.Length))   geom_beeswarm(side = 1L) 
#> Warning: Ignoring unknown parameters: side

However using the {see} package you can get pretty close to what I think you're asking for.

library(see)
# this works if you turn off the violin half
ggplot(iris,aes(Species, Sepal.Length))  
  geom_violindot(fill = NA, color = NA, color_dots = "black", size_dots = 1, fill_dots = "black")

^{Created on 2022-02-08 by the reprex package (v2.0.1)}