CODEWITHSUNDEEP
rggplot2
ljwharbers
ljwharbers

Reputation: 393

Is there a geom for the ordering of points within a variable?

I was wondering if there is a certain geom or function to make dot/jitter plots where the dots are ordered based on their value. See example below:

enter image description here

Upvotes: 1

Views: 57

Answers (2)

Allan Cameron
Allan Cameron

Reputation: 173793

A bit late to the party, but the alternative way would be to use facets:

df <- data.frame(x = rep(LETTERS[1:20], 50), y = rnorm(1000))

library(dplyr)
library(ggplot2)

df %>% 
  group_by(x) %>%
  mutate(new_x = seq_along(x),
         y = sort(y)) %>%
  ggplot(aes(new_x, y)) +
  geom_rect(aes(fill = as.factor(as.numeric(as.factor(x)) %% 2)),
            xmin = -Inf, xmax = Inf, ymin = -Inf, ymax = Inf) +
  geom_point(color = "red4", size = 0.5) +
  scale_fill_manual(values = c("white", "gray90"), guide = guide_none()) +
  facet_grid(.~x, switch = "x") +
  theme_classic() +
  theme(panel.spacing = unit(0, "points"),
        strip.background = element_blank(),
        axis.text.x = element_blank(),
        axis.ticks.length.x = unit(0, "points"),
        strip.placement = "outside")

enter image description here

Upvotes: 2

teunbrand
teunbrand

Reputation: 37923

I don't think there exists one, but we could write a quick and dirty position function that would transform the data in the appropriate way. Example below:

library(tidyverse)

df <- data.frame(
  var = LETTERS[sample(1:2, 100, TRUE)],
  val = rnorm(100)
)

position_orderdodge <- function(
  width = NULL
) {
  ggproto(NULL, PositionOrderdodge, width = width)
}

PositionOrderdodge <- ggproto(
  "PositionOrderdodge", PositionDodge,
  compute_panel = function(data, params, scales) {
    data <- flip_data(data, params$flipped_aes)
    w <- params$width
    data <- data %>% group_by(x, group) %>%
      mutate(rank = scales::rescale(rank(y), to = c(-w, w)))
    data <- as.data.frame(data)
    data$x <- data$x + data$rank
    data$rank <- NULL
    flip_data(data, params$flipped_aes)
  }
)

ggplot(df, aes(var, val)) +
  geom_point(position = position_orderdodge(0.2))

Created on 2020-12-07 by the reprex package (v0.3.0)

Note: not tested beyond this example

Upvotes: 5

Related Questions