gg_jitterbox.R

# function to do a dodged half-boxplot and jittered points next to each other
#
# data_in should be a data frame
# factor_col should be a bare column name (not a string)
#     although it will work if that column is factor or a character type
# numeric_col is the y axis continuous variable
# offset is the width of the boxplots and jittered point cloud
#
# the basic approach is to draw a boxplot without the tails
#   (e.g. only the interquartile range) and then use segments to add the
#    min and max values
#
# in part based on ideas here:
#    https://groups.google.com/forum/#!topic/ggplot2/jAyiMHf7DPU
#    http://ggplot2.tidyverse.org/reference/geom_boxplot.html

#### NOTE
# Only works with older versions of ggplot, e.g. this won't work with ggplot2 version 3+ 
# library("devtools")
# install_version("ggplot2", version = "2.2.1", repos = "http://cran.us.r-project.org")

library("rlang")
library("dplyr")
library("ggplot2")

gg_jitterbox <- function(data_in, factor_col, numeric_col, offset) {

  # turn bare args into quosures
  quo_factor <- enquo(factor_col)
  quo_numeric <- enquo(numeric_col)

  # do the base R stuff that doesn't play nice with quosures
  # the extra factor() call deals with the factor_col parameter potentially
  # being character type - otherwise finding levels() etc will fail
  # quo_text(quo_factor) just gives back the string of the column name that we
  # put as a bare parameter
  numeric_factor <- as.numeric(factor(data_in[[quo_text(quo_factor)]]))
  ftr_breaks <- seq(length(levels(factor(data_in[[quo_text(quo_factor)]]))))
  ftr_labels <- levels(factor(data_in[[quo_text(quo_factor)]]))

  # easiest to do this first bit with dplyr instead of pulling out
  # of the ggplot object, because we then only have to call ggplot() once
  data_in %>%
    
    # another check to make sure we have factors when expected
    mutate_if(is.character, as.factor) %>%

    # !! unquotes quosures, but only works well with tidyverse
    group_by(!!quo_factor) %>%
    mutate(d_ymin = min(!!quo_numeric),
           d_ymax = max(!!quo_numeric),
           d_lower = quantile(!!quo_numeric, 0.25),
           d_middle = median(!!quo_numeric),
           d_upper = quantile(!!quo_numeric, 0.75)) %>%
    
    ggplot() +
      
      # aes_() requires quoted formula types, but allows mixing
      # of enquo-ed bare names (e.g. quo_factor) and just
      # normal dplyr-style column names
      geom_boxplot(aes_(x = ~numeric_factor - offset,
                        ymin = ~d_lower,
                        ymax = ~d_upper,
                        lower = ~d_lower,
                        middle = ~d_middle,
                        upper = ~d_upper,
                        width = 2 * offset,
                        fill = quo_factor),
                   stat = "identity") +
      
      geom_jitter(aes_(x = ~numeric_factor + offset,
                       y = quo_numeric,
                       color = quo_factor),
                  width = offset - 0.25 * offset,
                  height = 0) +
      
      # bottom vertical segment
      geom_segment(aes(x = numeric_factor,
                        y = d_ymin,
                        xend = numeric_factor,
                        yend = d_lower)) +
      
      # top vertical segment
      geom_segment(aes(x = numeric_factor,
                        y = d_ymax,
                        xend = numeric_factor,
                        yend = d_upper)) +
      
      # top horizontal segment
      geom_segment(aes(x = numeric_factor - offset,
                        y = d_ymax,
                        xend = numeric_factor,
                        yend = d_ymax)) +
      
      # top vertical segment
      geom_segment(aes(x = numeric_factor - offset,
                        y = d_ymin,
                        xend = numeric_factor,
                        yend = d_ymin)) +
      
      # have to manually add in the x scale because we made everything numeric
      # to do the shifting
      scale_x_continuous(breaks = ftr_breaks,
                         labels = ftr_labels) +
      
      # this also needs to be added manually because of the change to numeric
      labs(x = quo_text(quo_factor))

}

# example with factor column stored as factor
gg_jitterbox(data_in = iris,
             factor_col = Species,
             numeric_col = Sepal.Length,
             offset = 0.1) +
  ggtitle("Iris and some Quo-Fu")

# example with factor column stored as character
gg_jitterbox(data_in = mpg,
             factor_col = manufacturer,
             numeric_col = hwy,
             offset = 0.2) +
  ggtitle("mpg and some Quo-Fu")