Skip to contents

Halftone fills for journal figures

Source: vignettes/gghalftone.Rmd
gghalftone.Rmd

A halftone prints tone as dots of varying size on a fixed lattice. gghalftone applies this to ggplot2 fills at draw time. The lattice has a physical pitch in millimetres, so a figure saved at 89 mm and at 183 mm gets the same screen.

Use it for journal figures. A confidence band stays legible in greyscale, overlapping intervals do not hide one another, and a bar chart can be decoded from its hatching alone.

Screen a whole plot

halftone_plot() takes a finished plot and returns it printed. It screens every filled layer, draws a paper hairline under every line that crosses a screen, and adds the theme modifier. Any argument of with_halftone() passes through.

m <- loess(dist ~ speed, cars, span = 0.9)
nd <- data.frame(speed = seq(4, 25, length.out = 80))
pr <- predict(m, nd, se = TRUE)
nd$fit <- pr$fit; nd$lo <- pr$fit - 1.96 * pr$se.fit; nd$hi <- pr$fit + 1.96 * pr$se.fit

p <- ggplot(nd, aes(speed)) +
  geom_ribbon(aes(ymin = lo, ymax = hi), fill = "steelblue") +
  geom_line(aes(y = fit), linewidth = 0.6) +
  geom_point(data = cars, aes(speed, dist), size = 1) +
  labs(x = "Speed (mph)", y = "Stopping distance (ft)") + theme_classic(base_size = 8)

halftone_plot(p)

The treatment follows the geom. Ribbons, areas, densities, bars, tiles, polygons, sf geometries, violins, boxplots and smooths are screened. Lines, paths, steps, contours and points are haloed, but only where a screen sits under them. Text, error bars, rugs and reference lines are left alone, and so is any geom it does not recognise.

It screens the fill and nothing else, so line weights, point shapes and fill colours stay as you set them. The rest of this vignette composes the same figure layer by layer.

Screen a confidence band

with_halftone() wraps any filled layer. The layer draws as usual, and its fill is replaced by a screen in the fill colour, clipped to the shape.

m <- loess(dist ~ speed, cars, span = 0.9)
nd <- data.frame(speed = seq(4, 25, length.out = 80))
pr <- predict(m, nd, se = TRUE)
nd$fit <- pr$fit; nd$lo <- pr$fit - 1.96 * pr$se.fit; nd$hi <- pr$fit + 1.96 * pr$se.fit

ggplot(nd, aes(speed)) +
  with_halftone(geom_ribbon(aes(ymin = lo, ymax = hi), fill = halftone_inks[["blue"]])) +
  with_halo(geom_line(aes(y = fit), colour = halftone_inks[["blue"]], linewidth = 0.35)) +
  geom_point(data = cars, aes(speed, dist), shape = 21, fill = "white", size = 0.8, stroke = 0.3) +
  labs(x = "Speed (mph)", y = "Stopping distance (ft)") +
  theme_classic(base_size = 8) + theme_halftone()

Three package defaults apply here:

  • The band’s tone follows the likelihood of the estimate across the interval. Tone is full on the fitted line and 0.146 at the 95% limit.
  • with_halo() draws a 0.09 mm paper stroke under the line, so the line stays readable where it crosses dots.
  • theme_halftone() adds only what a halftone needs to the theme it follows: a paper ground with no gridlines under the screen, legend keys large enough to show a screen, and the ink palette. Sizes and fonts stay with theme_classic().

Plot survival with three strata

The data are the North Central Cancer Treatment Group lung cancer trial (Loprinzi et al. 1994), distributed as survival::lung.

Groups that share a panel are printed on one lattice. Where two bands overlap, their dots alternate, so both inks stay visible.

library(survival)
fit <- survfit(Surv(time, status) ~ ph.ecog, data = subset(lung, ph.ecog < 3))
s <- km_steps(fit); levels(s$strata) <- paste("ECOG", levels(s$strata))
cens <- km_censor(fit); levels(cens$strata) <- levels(s$strata)

ggplot(s, aes(time, group = strata)) +
  with_halftone(geom_ribbon(aes(ymin = lo, ymax = hi, fill = strata))) +
  with_halo(geom_step(aes(y = surv, colour = strata), linewidth = 0.35)) +
  geom_point(data = cens, aes(time, surv, colour = strata), shape = "|", size = 1.6) +
  scale_y_continuous(labels = scales::percent) +
  guides(colour = "none") +
  labs(x = "Days since diagnosis", y = "Overall survival", fill = NULL) +
  theme_classic(base_size = 8) + theme_halftone() + theme(legend.position = "inside", legend.position.inside = c(0.82, 0.86))

km_steps() converts a survfit object into the step-shaped frame that the ribbon needs. km_censor() returns the censor marks and km_risk() the number at risk. On ggplot2 4.0 the modifier sets the ink palette as the default.

Screen bars

For geoms whose groups tile the plane (bars, areas, polygons, tiles), each fill also gets its own screen angle and shape. The legend keys show both, and the figure survives greyscale printing.

d <- data.frame(g = factor(c("BOS", "RAS", "Mixed", "Undef."), c("BOS", "RAS", "Mixed", "Undef.")), n = c(52, 21, 14, 13))
ggplot(d, aes(g, n, fill = g)) +
  with_halftone(geom_col(width = 0.7)) +
  scale_y_continuous(expand = expansion(c(0, 0.08))) +
  labs(x = NULL, y = "Patients (%)", fill = NULL) +
  theme_classic(base_size = 8) + theme_halftone()

For a one-ink figure, map the screen aesthetic and set shape = "line".

ggplot(d, aes(g, n, screen = g)) +
  with_halftone(geom_col(width = 0.7, fill = "black", colour = "black", linewidth = 0.25), shape = "line") +
  scale_screen_discrete(guide = "none") +
  scale_y_continuous(expand = expansion(c(0, 0.08))) +
  labs(x = NULL, y = "Patients (%)") +
  theme_classic(base_size = 8) + theme_halftone()

Save

ggsave_journal() saves at a journal column width. The file extension selects the format: PNG for proofs, TIFF for submission, PDF for a vector file in which every dot is a path.

ggsave_journal("fig1.pdf", p, "single", height = 62)
ggsave_journal("fig1.tiff", p, "double", height = 90)
halftone_proof(p)   # final-size render plus a 4x crop of the panel centre

Defaults

  • Pitch is 0.35 mm (73 lines per inch). A coarser pitch reads as a dot pattern, a finer one as a flat tint.
  • No feature is smaller than 0.09 mm (0.25 pt), the minimum line weight in journal artwork guidelines (Nature Portfolio; Elsevier). Light tone becomes sparse dots at the minimum size.
  • The tone profile follows the geometry: flat where the interior is the value (bars, areas, maps), likelihood for intervals, a soft vignette for densities.
  • One ink register applies across figure types, so a set of figures looks consistent.
  • Overlapping groups share one lattice. A fill’s alpha becomes screen coverage, and the output holds no partial transparency.

References