Awesome

ggpointdensity

Introduces geom_pointdensity(): A cross between a scatter plot and a 2D density plot.

Installation

To install the package, type this command in R:

install.packages("ggpointdensity")

# Alternatively, you can install the latest
# development version from GitHub:
if (!requireNamespace("devtools", quietly = TRUE))
    install.packages("devtools")
devtools::install_github("LKremer/ggpointdensity")

Motivation

There are several ways to visualize data points on a 2D coordinate system: If you have lots of data points on top of each other, geom_point() fails to give you an estimate of how many points are overlapping. geom_density2d() and geom_bin2d() solve this issue, but they make it impossible to investigate individual outlier points, which may be of interest.

geom_pointdensity() aims to solve this problem by combining the best of both worlds: individual points are colored by the number of neighboring points. This allows you to see the overall distribution, as well as individual points.

Changelog

Added method argument and renamed the n_neighbor stat to density. The available options are method="auto", method="default" and method="kde2d". default is the regular n_neighbor calculation as in the CRAN package. kde2d uses 2D kernel density estimation to estimate the point density (credits to @slowkow). This method is slower for few points, but faster for many (ca. >20k) points. By default, method="auto" picks either kde2d or default depending on the number of points.

Demo

Generate some toy data and visualize it with geom_pointdensity():

library(ggplot2)
library(dplyr)
library(viridis)
library(ggpointdensity)

dat <- bind_rows(
  tibble(x = rnorm(7000, sd = 1),
         y = rnorm(7000, sd = 10),
         group = "foo"),
  tibble(x = rnorm(3000, mean = 1, sd = .5),
         y = rnorm(3000, mean = 7, sd = 5),
         group = "bar"))

ggplot(data = dat, mapping = aes(x = x, y = y)) +
  geom_pointdensity() +
  scale_color_viridis()

Each point is colored according to the number of neighboring points. (Note: this here is the dev branch, where I decided to plot the density estimate instead of n_neighbors now.) The distance threshold to consider two points as neighbors (smoothing bandwidth) can be adjusted with the adjust argument, where adjust = 0.5 means use half of the default bandwidth.

ggplot(data = dat, mapping = aes(x = x, y = y)) +
  geom_pointdensity(adjust = .1) +
  scale_color_viridis()
 
ggplot(data = dat, mapping = aes(x = x, y = y)) +
  geom_pointdensity(adjust = 4) +
  scale_color_viridis()

Of course you can combine the geom with standard ggplot2 features such as facets...

# Facetting by group
ggplot(data = dat, mapping = aes(x = x, y = y)) +
  geom_pointdensity() +
  scale_color_viridis() +
  facet_wrap( ~ group)

... or point shape and size:

dat_subset <- sample_frac(dat, .1)  # smaller data set
ggplot(data = dat_subset, mapping = aes(x = x, y = y)) +
  geom_pointdensity(size = 3, shape = 17) +
  scale_color_viridis()

Zooming into the axis works as well, keep in mind that xlim() and ylim() change the density since they remove data points. It may be better to use coord_cartesian() instead.

ggplot(data = dat, mapping = aes(x = x, y = y)) +
  geom_pointdensity() +
  scale_color_viridis() +
  xlim(c(-1, 3)) + ylim(c(-5, 15))

ggplot(data = dat, mapping = aes(x = x, y = y)) +
  geom_pointdensity() +
  scale_color_viridis() +
  coord_cartesian(xlim = c(-1, 3), ylim = c(-5, 15))

Authors

Lukas PM Kremer (@LPMKremer) and Simon Anders (@s_anders_m), 2019