Annotation & Storytelling

The difference between an exploratory plot and a publication plot is the annotation. A well-annotated graph answers the reader’s questions on the graph itself: What am I looking at? Which point matters? What happened here? Tufte’s principle applies: write the labelling directly on the graph.

Code 
library(ggplot2)
library(dplyr)
library(gapminder)

gap2007 <- gapminder %>% filter(year == 2007)

Direct labels instead of legends

Legends force the reader’s eye to travel back and forth. Where possible, label the data directly.

Code 
library(ggrepel)

# Label only a meaningful subset — labelling everything is labelling nothing
highlight <- gap2007 %>%
  filter(country %in% c("Norway", "Germany", "China", "India",
                        "Nigeria", "Haiti", "Afghanistan"))

ggplot(gap2007, aes(x = gdpPercap, y = lifeExp)) +
  geom_point(alpha = 0.3, color = "grey50") +
  geom_point(data = highlight, color = "darkred", size = 2.5) +
  geom_text_repel(data = highlight, aes(label = country), size = 4) +
  scale_x_log10() +
  labs(x = "GDP per Capita (log scale)", y = "Life Expectancy") +
  theme_classic() +
  theme(axis.text = element_text(size = 14),
        axis.title = element_text(size = 15))

geom_text_repel() (from the ggrepel package) automatically moves labels so they don’t overlap points or each other. Plain geom_text() will happily print labels on top of your data.

Ending line charts with the series name replaces the legend entirely:

Code 
d <- gapminder %>% filter(country %in% c("Germany", "India", "China"))

ggplot(d, aes(x = year, y = lifeExp, color = country)) +
  geom_line(linewidth = 1.1) +
  geom_text_repel(
    data = d %>% group_by(country) %>% filter(year == max(year)),
    aes(label = country),
    direction = "y", hjust = 0, nudge_x = 1, size = 4.5
  ) +
  scale_color_viridis_d(option = "B", end = 0.85, guide = "none") +
  scale_x_continuous(limits = c(1952, 2015)) +
  labs(x = NULL, y = "Life Expectancy") +
  theme_classic() +
  theme(axis.text = element_text(size = 14),
        axis.title = element_text(size = 15))

Highlighting: one series in focus, the rest as context

gghighlight greys out everything except what you want the reader to see. The “compared to what?” question answered visually.

Code 
library(gghighlight)

ggplot(gapminder, aes(x = year, y = lifeExp, group = country)) +
  geom_line(color = "darkred", linewidth = 1) +
  gghighlight(country %in% c("Rwanda", "Cambodia"),
              unhighlighted_params = list(color = "grey85", linewidth = 0.3)) +
  labs(x = NULL, y = "Life Expectancy",
       title = "Life expectancy collapses during mass violence",
       subtitle = "All other countries shown in grey") +
  theme_classic() +
  theme(axis.text = element_text(size = 14),
        axis.title = element_text(size = 15))

Annotating events and reference values

annotate() adds one-off text, arrows, and shading without needing a data frame. Use it to mark events, thresholds, and periods.

Code 
germany <- gapminder %>% filter(country == "Germany")

ggplot(germany, aes(x = year, y = lifeExp)) +
  annotate("rect", xmin = 1961, xmax = 1989, ymin = -Inf, ymax = Inf,
           alpha = 0.1, fill = "steelblue") +
  annotate("text", x = 1975, y = 78, label = "Berlin Wall era",
           color = "steelblue", size = 4.5) +
  geom_line(linewidth = 1.1) +
  geom_hline(yintercept = 75, linetype = "dashed", color = "grey40") +
  annotate("text", x = 1955, y = 75.7, label = "75 years",
           color = "grey40", hjust = 0, size = 4) +
  labs(x = NULL, y = "Life Expectancy") +
  theme_classic() +
  theme(axis.text = element_text(size = 14),
        axis.title = element_text(size = 15))

Titles that state the finding

Compare “Life Expectancy by Continent” with “Africans live 15 years less than Europeans on average”. The first describes the axes; the second states the finding. Newspapers do this well; academic figures can too (use the caption if the title must stay neutral).

A useful division of labour:

  • Title: the finding, in one sentence.
  • Subtitle: the data and unit of analysis.
  • Caption: source and notes (labs(caption = "Data: Gapminder, 2007")).

Checklist before you export

  1. Can a reader understand the plot without the surrounding text?
  2. Is everything the reader must compare labelled directly?
  3. Do title/subtitle state what to see, not just what is plotted?
  4. Is the source on the graph?
  5. Have you erased annotation ink that isn’t doing work? (Annotation is data-ink only when it carries information.)