In jrnold/ggtufte: Themes and geoms implementing Tufte's work in 'ggplot2'

library(tufte)
# invalidate cache when the tufte version changes
knitr::opts_chunk$set(tidy = FALSE, cache.extra = packageVersion('tufte'))
options(htmltools.dir.version = FALSE)

Introduction

library("ggplot2")
library("ggtufte")
library("scales")
extrafont::loadfonts(quiet = TRUE)

Scatter plots

p <- ggplot(mtcars, aes(wt, mpg)) +
  geom_point()

p + theme_bw()

The ggplot2 theme provided with this package, theme_tufte, removes most of the ink from the plot. There are no grid-lines and no borders or axis.

p + theme_tufte()

Instead of using axis lines the ink-to-data ratio can be improved by using the axis lines to encode the range of the data. The geom_rangeframe() will add lines representing the range of the data along the axes.

p + theme_tufte() +
  geom_rangeframe()

However, the axis breaks do not match up with the range. It would be more informative to label the minimum and maximum values of breaks.

p + theme_tufte() +
  geom_rangeframe() +
  scale_y_continuous(breaks = range_breaks(),
                     labels = format_format(digits = 1)) +
  scale_x_continuous(breaks = range_breaks(),
                     labels = format_format(digits = 1))

p + theme_tufte() +
  geom_rangeframe() +
  scale_y_continuous(breaks = range_breaks()(mtcars$mpg),
                     labels = format_format(digits = 1)) +
  scale_x_continuous(breaks = range_breaks()(mtcars$wt),
                     labels = format_format(digits = 1))

p + theme_tufte(ticks = FALSE) +
  geom_rug() +
  scale_y_continuous(breaks = range_breaks()(mtcars$mpg),
                     labels = format_format(digits = 1)) +
  scale_x_continuous(breaks = range_breaks()(mtcars$wt),
                     labels = format_format(digits = 1))

Quartile Plots

p <- ggplot(mpg, aes(class, hwy)) +
  theme(axis.ticks.x = element_blank())

Tufte Ch. 6 proposes several redesigns of the venerable box plot. Unlike the Tukey's boxplot, these box plots, display the five-number summary (median, 25% quartile, 75% quartile, maximum, minimum), instead of the stats and outliers calculated in Tukey's boxplot.

This displays the standard box plot in ggplot2.

p + geom_boxplot(stat = StatFivenumber)

I use the StatFivenumber stat to calculate the fiver-number summary rather than the summary used in the boxplot.

@Tufte calls his redesign a quartile-plot. The function geom_quartileplot() implements these.

The first redesign erases the box from the boxplot. The whiskers remain, negative space represents the interquartile and a point represents the median.

p + geom_quartileplot()

A second alternative is to invert the first. Points are used to represent the minimum and maximum, a line for the interquartile range, and a gap in the line to represent the median.

p + geom_quartileplot(whisker.line = FALSE, middle.line = TRUE)

A third alternative is to use a line to represent the range, and a gap to represent the median. The interquartile range is represented by a thicker section of line.

p + geom_quartileplot(whisker.line = TRUE, middle.line = TRUE,
                      hoffset = 0, fatten = 2)

A fourth alternative uses lines for the upper and lower whiskers. The interquartile range is represented by a line offset from the first. The median is still represented by a gap.

p + geom_quartileplot(whisker.line = TRUE, middle.line = TRUE)

Theme

p <- ggplot(mtcars, aes(x = wt, y = mpg, colour = factor(cyl))) +
  geom_point() + 
  facet_wrap(~ am) +
  labs(title = "Title", subtitle = "Subtitle", caption = "Caption")
p + theme_tufte()