Appendix C — Figure Conventions

---

A default R plot is rarely a good plot, and the gap between the two is mostly a handful of arguments. This appendix collects the choices the book makes and explains what each one is for, so you can apply them to figures the book never draws. Most of them are house style rather than rules: another book would pick different colors and be no worse for it. The one section that is not a matter of taste is Accessibility, which is about whether a reader can use your figure at all. The theme running through the rest is that ink should carry information, since borders, bold titles, and full-strength colors all draw the eye without telling the reader anything. Most of the code blocks here are shown for reference and are not run, but the two worked examples are.

C.1 Default Versus House Style

The two panels below plot the same data. Only the arguments differ.

X <- USArrests[,'Murder']
par(mfrow=c(1,2), mar=c(4, 4, 2, 1))

# Default settings
hist(X)

# House style
hist(X, breaks=15, border=NA, freq=FALSE, main=NA,   # fewer bins, since n is only 50
    xlab='Murder arrests (per 100k)')
abline(v=mean(X), col=rgb(1, 0, 0, .8), lwd=2)
title(paste0('mean= ', round(mean(X), 2)), font.main=1)

Four things changed. The bar borders are gone, so the shape of the distribution reads as one silhouette instead of twenty outlined boxes. The vertical axis is a density rather than a count, which makes the histogram comparable to a fitted curve. The axis label says what the variable is and in what units. The title reports a number rather than repeating the variable name that the axis label already gave.

C.2 Histograms

hist(X, breaks=25, border=NA, freq=FALSE, main=NA, xlab='label')

Argument	Convention	Notes
border	NA	Never use visible borders on bars
freq	FALSE	Density scale, not counts
main	NA or omit	Add titles via title() instead
breaks	20–60 integer	Adjust to sample size; 25 is a good default
col	Default grey, or grey(0, alpha)	Use transparency for overlays
xlab	Short descriptive string	Use expression() for math notation

The density scale is the argument that matters most often. With freq=FALSE the bars integrate to one, so you can overlay a theoretical density or a kernel estimate on the same axes. With the default freq=TRUE you cannot, because the two are on different scales.

C.3 Scatterplots

plot(y ~ x, pch=16, col=grey(0, .5), data=dat, main=NA, xlab='x', ylab='y')

Argument	Convention	Notes
pch	16 (solid circle)	Standard throughout the book
col	grey(0, alpha)	Alpha in [0.05, 0.5] depending on \(n\)
main	NA or omit	Titles via title() if needed
cex	Reduce for large \(n\) (e.g., .5)	Default is fine for small datasets

Transparency is how a scatterplot survives a large sample. Opaque points overplot into a solid blob that hides where the data are dense, while transparent points darken where observations pile up, so the plot doubles as a picture of the joint density.

Sample size	Suggested alpha
\(n < 100\)	0.5
100–1000	0.25
1000–10000	0.1
\(n > 10000\)	0.05

C.4 Titles and Colors

Add titles with title() after the plot call, rather than with main= inside it.

title(paste0('mean= ', round(m, 2)), font.main=1)

Always set font.main=1, which is plain text. R’s default is bold, and a bold title competes with the data for attention. Keep the text short, usually a statistic or a single phrase, and use title('label', outer=TRUE) for a label spanning a multi-panel figure.

The Grey and RGB System.

Scatterpoints and histogram fills use grey() with a first argument of 0, varying only the transparency. Anything you want the reader to look at uses rgb() with an explicit alpha channel.

Code	Color	Use for
rgb(1, 0, 0, .8)	Red	Means, fitted lines, key statistics
rgb(0, 0, 1, .8)	Blue	SD bounds, secondary reference lines
rgb(0, 0, 0, .8)	Black	Horizontal and vertical reference lines

Use rgb() rather than numeric codes such as col=2 or names such as 'red'. The numeric codes depend on the current palette, so the same number can produce a different color in a different session, and neither form lets you set transparency.

For two groups, use cols <- c(rgb(.8, 0, 0, .5), rgb(0, 0, .8, .5)). For three or more, use cols <- hcl.colors(k, alpha=.45), where k is the number of groups. hcl.colors() spaces colors evenly in a perceptual space, so no one group looks louder than the others.

C.5 Accessibility

Everything above this section is a preference. This section is not. Around one man in twelve has some form of color vision deficiency, your figure may be printed in greyscale or read on a washed-out projector, and a reader using a screen reader gets only the text you attach to it. Four habits cover most of the problem.

Never let color be the only difference between groups. Pair it with a shape for points (pch) or a line type for lines (lty), so the groups stay separable when the color does not survive. Base R also ships a colorblind-safe categorical palette, which is a better starting point than picking hues yourself.

palette.colors(4, 'Okabe-Ito')
## [1] "#000000" "#E69F00" "#56B4E9" "#009E73"

Keep text large enough to survive scaling. A figure is drawn at one size and displayed at another, usually smaller, so text set below the default can become unreadable. Set cex.lab and cex.axis at 1 or above, and never shrink text that is words rather than plotting symbols.

Caption the figure. A caption states what the reader is looking at, so the figure can be understood without hunting for the paragraph that introduced it. In a Quarto or R Markdown report, fig-cap puts it under the figure and numbers it for you.

Write alt text that gives the finding. Alt text replaces the figure for a reader who cannot see it, so describing the ink (“a scatterplot with red and blue points”) is useless. Say what the figure shows. fig-alt sets it.

The example below does all four. Group membership is carried by shape as well as color, the label text is enlarged, and the caption and alt text are attached to the chunk rather than drawn onto the plot.

high <- USArrests[,'Assault'] > median(USArrests[,'Assault'])

plot(USArrests[,'Murder'] ~ USArrests[,'UrbanPop'],
    pch=ifelse(high, 16, 17),                                  # shape carries the group
    col=ifelse(high, rgb(.8, 0, 0, .6), rgb(0, 0, .8, .6)),    # color repeats it
    cex.lab=1.1, cex.axis=1.1,
    xlab='Urban population (%)', ylab='Murder arrests (per 100k)')
legend('topleft', legend=c('High assault', 'Low assault'),
    pch=c(16, 17), col=c(rgb(.8, 0, 0, .6), rgb(0, 0, .8, .6)), bty='n')

Murder arrests against urban population, split at the median assault rate.

Note that the legend repeats both pch and col. A legend that gives only the colors defeats the point of encoding the group twice.

To check a finished figure, view it in greyscale. If you cannot tell the groups apart, the shapes are not doing their job.

C.6 Reference Lines and Annotations

# Vertical line at a statistic (e.g., mean)
abline(v=value, col=rgb(1, 0, 0, .8), lwd=2)

# Horizontal reference (e.g., zero line)
abline(h=0, col=rgb(0, 0, 0, .8), lty=2)

# Confidence interval bounds
abline(v=ci_bounds, col=rgb(0, 0, 1, .8), lty=2)

Element	col	lwd	lty
Sample statistic	rgb(1, 0, 0, .8)	2	1
CI or theoretical value	rgb(0, 0, 1, .8)	1	2
Subtle reference	grey(0.5)	1	3

The pattern is that solid and thick means “computed from this sample”, while dashed and thin means “a reference to compare it against”. A reader who has seen one figure in the book can then read the next one without a legend.

For text annotations, match the color to the element being labeled and use expression() or bquote() for math.

text(x, y, label, col=rgb(0, 0, 1, .8), adj=0)

Use adj=0 for left-aligned text and adj=0.5 for centered.

Fitted and Overlay Lines.

lines(x_sorted, y_hat, col=rgb(1, 0, 0, .8), lwd=2)

Draw fitted lines with lines() rather than curve(), using red at lwd=2 for the primary fit and blue at lwd=1, lty=2 for a comparison fit. lines() takes the fitted values you actually computed, whereas curve() re-evaluates a function, which can quietly differ from the model you fit. Sort by the horizontal variable first, or lines() will zigzag back and forth across the plot.

C.7 Layouts and Legends

par(mfrow=c(rows, cols))

Use par(mfrow=c(1,2)) for two panels side by side, c(1,3) for three, and c(2,2) for a grid. Set small top and right margins with par(mar=c(4, 4, 1, 1)). If you change par() inside a function, restore it on exit.

op <- par(no.readonly=TRUE)
on.exit(par(op), add=TRUE)

legend('topright', legend=c('A', 'B'),
    col=c(rgb(1, 0, 0, .8), rgb(0, 0, 1, .8)), lwd=2, cex=.8)

Position legends with a string such as 'topright' or 'topleft' rather than coordinates, so the legend moves with the data instead of landing on top of it. Add bty='n' to drop the box when it clutters the plot, keep entries to one to three words, and match lty, pch, and col to the plot symbols.

Boxplots.

boxplot(X1, X2, names=c('A', 'B'), main=NA, xlab='Groups')

Use main=NA and add titles with title(..., font.main=1). Color grouped boxplots with col=hcl.colors(k, alpha=.45), and set border='white' or border=NA when a fill color is specified.

C.8 Reference Card

House style, before you save a figure:

• histograms use border=NA
• density histograms use freq=FALSE
• scatterplots use pch=16
• points use grey(0, alpha) with alpha matched to \(n\)
• any title sets font.main=1
• emphasis colors use rgb() with alpha, not col=2 or 'red'
• multi-panel layouts use par(mfrow=...), not layout()

Accessibility, before anyone else sees it:

• both axes are labeled, in words, with units
• groups differ by pch or lty, not by color alone
• the legend repeats every channel used, not just color
• cex.lab and cex.axis are at 1 or above
• it still reads in greyscale
• a report figure has fig-cap, and fig-alt that states the finding

Further Reading.

• https://clauswilke.com/dataviz/ – Fundamentals of Data Visualization, free online, on which figures work and why.
• Tufte (2001) – the source of the ink-per-unit-information argument above.
• Data Analysis – applies these conventions to a full worked figure.

Tufte, Edward. 2001. The Visual Display of Quantitative Information. 2nd ed. Graphics Press. https://www.google.de/books/edition/The_Visual_Display_of_Quantitative_Infor/qmjNngEACAAJ.