nevrome / ggpointgrid

R Package - Rearrange scatter plot points on a regular grid

ggpointgrid

This package provides geoms to rearrange scatter-plot coordinates on regular grids while strictly avoiding over-plotting. The applications are similar to geom_jitter.

Installation

You can install the development version from github with the following command (in your R console):

if(!require('remotes')) install.packages('remotes')
remotes::install_github("nevrome/ggpointgrid")

Examples

library(ggplot2)
set.seed(5)

geom_pointgrid

geom_pointgrid aims to optimize the arrangement of observations on a regular grid. This works well for figures with continuously scaled x- and y-axes, so for scatter-plots or even map plots. Just as in geom_jitter the rearrangement of the points reduces accuracy and precision of positional information on x and y in favour of making every observation visible.

The grid properties are controlled with the parameters grid_x and grid_y, which allow to precisely specify the desired graticules.

df <- tibble::tibble(
  x = rep(c(1,1,2,3,3), times = 10),
  y = rep(c(1,3,2,1,3), times = 10),
  var = sample(c("A", "B", "C"), size = 50, replace = T)
)

coord <- coord_fixed(xlim = c(0.5,3.5), ylim = c(0.5,3.5))

p1 <- ggplot(df) +
  geom_point(aes(x, y, color = var)) +
  coord + ggtitle("geom_point")

p2 <- ggplot(df) +
  geom_jitter(aes(x, y, color = var), width = 0.3, height = 0.3) +
  coord + ggtitle("geom_jitter")

p3 <- ggplot(df) +
  ggpointgrid::geom_pointgrid(aes(x, y, color = var), grid_x = 15, grid_y = 15) +
  coord + ggtitle("geom_pointgrid")

p4 <- ggplot(df) +
  ggpointgrid::geom_pointgrid(
    aes(x, y, color = var),
    grid_x = seq(min(df$x) - 0.3, max(df$x) + 0.3, length.out = 18),
    grid_y = seq(min(df$y) - 0.3, max(df$y) + 0.3, length.out = 18)
  ) +
  coord + ggtitle("geom_pointgrid, grid specified")

cowplot::plot_grid(p1, p2, p3, p4)

geom_textgrid performs the same arrangement operation on text data. It is to geom_text what geom_pointgrid is to geom_point.

geom_pointrect

geom_pointrect was designed for a slightly different use-case than geom_pointgrid. Here all observations that share the x- and y-coordinate are spread out into a rectangular grid, representing only this one position. This is especially useful, when the x- and y- axis are ordinally scaled.

The order within each rectangle can be set by the order of the input data.frame and the arguments scale_x and scale_y control the size of the per-position box. The arguments round_x and round_y allow to specify how data on continuously scaled x- and y-axes should be aggregated.

df <- tibble::tibble(
  x = rep(letters[c(1,1,2,3,3)], times = 10),
  y = rep(letters[c(1,3,2,1,3)], times = 10),
  var = sample(c("A", "B", "C"), size = 50, replace = T)
) |> dplyr::arrange(var)

coord <- coord_fixed()

p4 <- ggplot(df) +
  geom_point(aes(x, y, color = var)) +
  coord + ggtitle("geom_point")

p5 <- ggplot(df) +
  geom_jitter(aes(x, y, color = var), width = 0.3, height = 0.3) +
  coord + ggtitle("geom_jitter")

p6 <- ggplot(df) +
  ggpointgrid::geom_pointrect(aes(x, y, color = var)) +
  coord + ggtitle("geom_pointrect")

p7 <- ggplot(df) +
  ggpointgrid::geom_pointrect(
    aes(x, y, color = var),
    scale_x = 0.2,
    scale_y = 0.2
  ) +
  coord + ggtitle("geom_pointrect, scaling set")

cowplot::plot_grid(p4, p5, p6, p7)