Plotting data

ggplot, part 2

Roland Krause

Rworkshop

Wednesday, 12 February 2025

Introduction

More ggplot2!

Scales catalogue

Scales for all

62 functions available!

Functions are based on this scheme:

scale_{aes}_{type}()

Arguments to change default

  • breaks, choose where are labels
  • labels, usually with package scales
  • trans, to change to i.e log

tidyr crossing + glue glue_data trick

To generate all combinations

tidyr::crossing(aes = c("fill", "colour", "alpha", "x", "y"),
         type = c("continuous", "discrete", "date")) |>
  glue::glue_data("scale_{aes}_{type}()")
scale_alpha_continuous()
scale_alpha_date()
scale_alpha_discrete()
scale_colour_continuous()
scale_colour_date()
scale_colour_discrete()
scale_fill_continuous()
scale_fill_date()
scale_fill_discrete()
scale_x_continuous()
scale_x_date()
scale_x_discrete()
scale_y_continuous()
scale_y_date()
scale_y_discrete()
scales::label_percent()(c(0.4, 0.6, .95))
[1] "40%" "60%" "95%"
scales::label_comma()(20^(1:4))
[1] "20"      "400"     "8,000"   "160,000"
scales::breaks_log()(20^(1:3))
[1]    10    30   100   300  1000  3000 10000

Scales

Scales transformation

Starwars: who is the massive guy?

ggplot(starwars, aes(x = height, y = mass)) +
  geom_point(size = 2) # from dplyr

ggplot(starwars, aes(x = height, y = mass)) +
  geom_point(alpha = 0.6, size = 2) +
  geom_text(data = \(x) filter(x, mass > 1000),
            aes(label = name), nudge_y = -0.2) +
  annotation_logticks(sides = "l") +
  scale_y_continuous(trans = "pseudo_log")

pseudo_log is available through scales.

All geometries geom_ can take either a data.frame or a function, here geom_text().

Native lambda, < 4.1 would be function(x) filter(x, mass > 1000).

Colors

Custom colors

Better: see Emil Hvitfeldt repo

Changing colours for categories

Default is scale_fill_hue()

filter(starwars, !is.na(gender)) |>
ggplot(aes(y = gender,
          fill = fct_lump_n(species, 8) |>
                 fct_infreq())) +
  geom_bar(position = "fill") +
  scale_fill_hue() +
  labs(fill = "top 8 Species\n(ranked)")

Rbrewer is a safe alternative

filter(starwars, !is.na(gender)) |>
ggplot(aes(y = gender,
          fill = fct_lump_n(species, 8) |>
                 fct_infreq())) +
  geom_bar(position = "fill") +
  scale_fill_brewer(palette = "Set1") +
  labs(fill = "top 8 Species\n(ranked)")

Predefined colour palettes

library(RColorBrewer)
par(mar = c(0, 4, 
            0, 0))
display.brewer.all()

Colour gradient, for continuous variables

The default gradient generated by ggplot2 is not very good… Better off using viridis (scale_color_viridis_c() c for continuous)

penguins |>
  ggplot(aes(x = bill_length_mm, 
             y = bill_depth_mm, 
             colour = body_mass_g)) +
  geom_point(alpha = 0.6, size = 5)

penguins |>
  ggplot(aes(x = bill_length_mm, y = bill_depth_mm, 
             colour = body_mass_g)) +
  geom_point(alpha = 0.6, size = 5) +
  scale_color_viridis_c()

Viridis palettes

  • 5 different scales
  • Also for discrete variables
  • viridis is colour blind friendly and nice in b&w
  • In ggplot2 since v3.0 but no the default 
penguins |>
  ggplot(aes(x = bill_length_mm, y = bill_depth_mm, 
             colour = species)) +
  geom_point(alpha = 0.6, size = 5) +
  scale_color_viridis_d()

Binning instead of gradient

Binning help grouping observations

  • Default blue gradient, viridis option
  • Number of bins, limits can be changed
penguins |>
  ggplot(aes(x = bill_length_mm, y = bill_depth_mm, 
             colour = body_mass_g)) +
  geom_point(alpha = 0.6, size = 5) +
  scale_color_binned(type = "viridis")

Merge similar guides

Size & colour on same variable ➡️ 2 guides

penguins |>
  ggplot(aes(x = bill_length_mm, y = bill_depth_mm, 
             colour = body_mass_g,
             size = body_mass_g)) +
  geom_point(alpha = 0.6) +
  scale_color_viridis_c()

guides() merge both aesthetics

penguins |>
  ggplot(aes(x = bill_length_mm, y = bill_depth_mm, 
             colour = body_mass_g, size = body_mass_g)) +
  geom_point(alpha = 0.6) +
  scale_color_viridis_c() +
  guides(colour = "legend")

Facets

ggplot(penguins, 
       aes(bill_depth_mm, 
           bill_length_mm)) +
  facet_grid(species ~ island)

Facets: facet_wrap()

Creating facets

  • Easiest way: facet_wrap()
  • Use a formula (in R ~)
  • facet_wrap() is for one var
  • Or the vars() function
ggplot(penguins, 
       aes(bill_depth_mm, 
           bill_length_mm)) +
  geom_point() +
  facet_wrap(~ species)

Facets layout

Specify the number of rows/columns:

  • ncol = integer
  • nrow = integer
fc <- ggplot(penguins, 
       aes(bill_depth_mm, 
           bill_length_mm)) +
  geom_point()
fc + facet_wrap(~ species, ncol = 2)

Facets, free scales

⚠️ Make comparison harder

  • x or y (free_x / free_y)
fc + facet_wrap(~ species, scales = "free_y")

  • Both axis
fc + facet_wrap(~ species, scales = "free")

facet_grid() to lay out panels in a grid

Specify a 2 sides formula

rows on the left, columns on the right separated by a tilde ~

ggplot(penguins, 
       aes(bill_depth_mm, 
           bill_length_mm)) +
  geom_point() +
  facet_grid(island ~ species)

Barplots in facets

facet_grid() can also be used with one variable, complemented by a placeholder: .

Waste of space

ggplot(starwars) +
  geom_bar(aes(y = gender, fill = sex)) +
  facet_grid(fct_lump_min(species, 4) ~ .) +
  labs(y = NULL)

Optimize by removing empty slots

ggplot(starwars) +
  geom_bar(aes(y = gender, fill = sex)) +
  facet_grid(fct_lump_min(species, 4) ~ .,
             space = "free", scales = "free_y") +
  labs(y = NULL)

Extensions

ggplot2 introduced the possibility for the community to create extensions, they are referenced on a dedicated site

(https://exts.ggplot2.tidyverse.org/gallery/)

Cosmetics 💅

Black theme

Using Bob Rudis hrbrthemes package

library(hrbrthemes)
ggplot(penguins,
       aes(x = bill_length_mm,
           y = bill_depth_mm,
           colour = species)) +
  geom_point() +
  geom_smooth(method = "lm", formula = 'y ~ x',
      # no standard error ribbon
              se = FALSE) +
  facet_grid(island ~ .) +
  labs(x = "length (mm)", y = "depth (mm)",
       title = "Palmer penguins",
       subtitle = "bill dimensions over location and species",
       caption = "source: Horst AM, Hill AP, Gorman KB (2020)") +
  # hrbrthemes specifications
  scale_fill_ipsum() +
  theme_ft_rc(14) +
  # tweak the theme
  theme(panel.grid.major.y = element_blank(),
   panel.grid.major.x = element_line(size = 0.5),
   plot.caption = element_text(face = "italic"),
   strip.text = element_text(face = "bold"),
   plot.caption.position = "plot")

Compose plots with patchwork

Patchwork is developed by Thomas Lin Pedersen, main maintainer of ggplot2.

Define 3 plots and assign them names

p1 <- ggplot(penguins, 
             aes(x = species)) +
  geom_bar() +
  labs(title = "Species distribution")
p2 <- ggplot(penguins, 
             aes(y = island)) +
  geom_bar() +
  labs(title = "Island but flipped")
p3 <- ggplot(penguins, 
             aes(x = body_mass_g,
                 y = bill_depth_mm,
                 colour = sex)) +
  geom_point()
p1

p2

p3

Now, compose them!

Compose plots with patchwork

patchwork provides an API using the classic arithmetic operators

library(patchwork)
(( p1 | p2 ) / p3) +
  # add tags and main title
  plot_annotation(tag_levels = 'A',
                  title = 'Plots about penguins') &
  # modify all plots recursively
  theme_minimal() +
  theme(text = element_text('Roboto'))

Polished example by Cédric Scherer

Using patchwork, ggforce T. Pedersen and ggtext Claus O. Wilke

Text paths, curved labels in data

library(geomtextpath)
ggplot(penguins,
       aes(x = bill_length_mm, y = bill_depth_mm, colour = species)) +
  geom_point(alpha = 0.3) +
  stat_ellipse(aes(label = species),
    geom = "textpath", hjust = c(0.55)) +
  theme_bw(14) + theme(legend.position = "none")

ggplot(penguins,
       aes(x = bill_length_mm / bill_depth_mm, 
           colour = species)) +
  stat_density(aes(label = species),
    geom = "textpath", hjust = c(0.25)) +
  theme_bw(14) + theme(legend.position = "none")

Exporting, interactive or passive mode

Right panel

  • Using the Export button in the Plots panel

ggsave

  • Save the ggplot object, 2nd argument, here p or last_plot()
  • Guesses the output type by the extension (jpg, png, pdf etc.)
ggsave("my_name.png", p, width = 60, height = 30, units = "mm")
ggsave("my_name.pdf", last_plot(), width = 50, height = 50, units = "mm")

Quarto documents

  • If needed, adjust the chunk options:
    • Size: fig.height, fig.width
    • Ratio: fig.asp…

Some tricks and mentions

Supplying colors as data

tribble(~color, ~border, ~x, ~y,
        "#295f58", "black", 2, 3,
        "#eee8a9", "black",  2, 5,
        "#00A4E1" , "#777", 4, 4,
        "#CC0000", "#777",  3, 2) |> 
  ggplot() +
  aes(x = x , y = y, 
      color = I(border),
      fill = I(color)) +
  geom_point(size = 12, shape = 21, stroke = 5) +
  labs(title = "I() to supply the identity") +
  coord_cartesian(xlim = c(1, 5), ylim = c(1,6))

Expanding coordinates with coord_cartesian()

Preferred, not manipulating the data, just changing the viewport.

Font sizes controlled through themes

ggplot(penguins) +
  aes(x = flipper_length_mm, y = body_mass_g, color = sex) +
  geom_point() +
  scale_color_manual(values = c("darkorange", "cyan4"), na.translate = FALSE) +
    labs(title = "Penguin flipper and body mass",
       caption = "Horst AM, Hill AP, Gorman KB (2020)",
       subtitle = "Dimensions for male/female Adelie, Chinstrap and Gentoo Penguins at Palmer Station LTER",
       x = "Flipper length (mm)",
       y = "Body mass (g)",
       color = "Penguin sex") +
  theme(plot.subtitle = element_text(size = 13),
        axis.title   = element_text(size = 11)) +
  theme(legend.position = "bottom",
        legend.background = element_rect(fill = "white", color = NA)) +
  theme(plot.caption = element_text(hjust = 0, face = "italic"),
        plot.caption.position = "plot")

Missing features

Geoms list here

  • geom_tile() heatmap
  • geom_bind2d() 2D binning
  • geom_abline() slope

Stats list here

  • stat_ellipse() (ggforge seen)
  • stat_summary() easy mean, 95CI etc.

Plot on multi-pages

  • ggforce::facet_grid_paginate() facets
  • gridExtra::marrangeGrob() plots

Customise theme elements

  • Legend & guide tweaks
  • Major/minor grids
  • Font, faces
  • Margins
  • Labels & ticks
  • Strip positions
  • See live examples of pre-built themes

Build your plot with point and click

Using esquisse Rstudio addin by dreamRs

Before we stop

You learned to:

  • Apprehend facets
  • Color schemes and gradients
  • Discover extensions
  • The endless potential of cosmetic improvements

Acknowledgments

Further reading

Thank you for your attention!

*: Palmer penguins, data are available by CC-0 license and Artwork by Allison Horst