Exercises

Download this quarto file with exercises. Solutions are on the Solutions page .

Code

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

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

🎨 Colors

1.1 Take this plot and give it (a) a monochromatic color scheme, (b) a complementary two-color scheme for highlighting Europe vs. all other continents. Use coolors.co or paletton.com to pick the values, and scale_color_manual() to apply them.

Code

ggplot(gap2007, aes(x = gdpPercap, y = lifeExp, color = continent)) +
  geom_point() +
  scale_x_log10()

1.2 Run colorblindr::cvd_grid() on your version of the plot from 1.1(b). Is your highlight still visible under deuteranopia? If not, fix it without changing the colors (hint: redundant encoding).

1.3 You want to map unemployment rates (0–25%) across German districts. Should the fill scale be sequential, diverging, or qualitative? What if you instead map change in unemployment (-5 to +5 percentage points)? Write down your answer, then create both scales with scale_fill_viridis_c() and scale_fill_gradient2() on any choropleth from the course.

📊 Chart Choice

2.1 For each question, name the most suitable chart type and the aesthetic mappings (x, y, color, size, facet), no code needed:

How is population distributed across countries in 2007?
Did the gap in life expectancy between Africa and Europe narrow between 1952 and 2007?
Which continent has the most within-continent inequality in GDP per capita?
Is there a relationship between population size and life expectancy?

2.2 Now implement two of your answers from 2.1 in ggplot2.

2.3 A colleague presents the market shares of 12 toothpaste brands as a pie chart. Explain, using the Cleveland & McGill results from the Data Visualization page, why this is hard to read and produce a better alternative with fake data (tibble(brand = LETTERS[1:12], share = ...)).

✍️ Exploring Gapminder

3.1 Create a histogram of gdpPercap for 2007. It will look skewed. Fix this with a log scale and explain in one sentence what a reader must now keep in mind.

3.2 Build a line plot of population over time for the five most populous countries of 2007. Replace the legend with direct labels at the end of each line (ggrepel).

3.3 Compute mean life expectancy per continent per year, and plot it as a line chart. Annotate the most striking feature of the plot with annotate().

🧠 Challenge: Explore a Development Pattern

Before seeing any example plots, try to visualize the following:

How does a country’s population size relate to life expectancy and income level?

Your task: Create a visualization using the gapminder dataset for the year 2007 that captures the relationship between:

lifeExp: Life expectancy
gdpPercap: GDP per capita
pop: Population

You may use any type of plot or mapping of variables (e.g. color, size), but aim to show all three variables clearly. Then annotate it so it works without any accompanying text: a title that states the finding, direct labels for at least three notable countries, and a source caption.

🔧 Redesign Gallery

Bad charts are the best teachers. For each of the following, (a) list what is wrong, (b) rebuild it properly.

4.1 The chartjunk special:

Code

gap2007 %>%
  arrange(desc(pop)) %>%
  slice(1:10) %>%
  ggplot(aes(x = country, y = pop, fill = country)) +
  geom_bar(stat = "identity", color = "black", linewidth = 1.5) +
  labs(title = "population of Countries in year 2007",
       x = "This is the X Axis", y = "Some Very Large Numbers") +
  theme(panel.grid.major = element_line(color = "purple", linewidth = 1),
        plot.background = element_rect(fill = "lightyellow"))

4.2 The dynamite plot:

Code

gap2007 %>%
  group_by(continent) %>%
  summarise(m = mean(lifeExp), se = sd(lifeExp) / sqrt(n())) %>%
  ggplot(aes(x = continent, y = m)) +
  geom_col(fill = "steelblue") +
  geom_errorbar(aes(ymin = m - se, ymax = m + se), width = 0.2)

Why is the bar misleading here? What does it hide? (Hint: what is the y-axis origin, and where did the raw data go?)

4.3 The truncated axis. Recreate any bar chart from the course, then add coord_cartesian(ylim = c(60, 85)). Describe the effect on a casual reader, and explain when axis truncation is legitimate (line charts?) and when it is a lie (bar charts?).

4.4 Find one real published chart (newspaper, report, social media) that violates at least one of Tufte’s principles from the Data Visualization page. Bring it to the next session with a one-paragraph critique or redesign it if the data is available. Sarah Leo’s Economist redesigns show the format.

🗺️ Bonus: Spatial

5.1 Using rnaturalearth and the 2007 Gapminder data (see Applied Examples), map population instead of GDP. Population is even more skewed than GDP what do you need to change about the fill scale to make the map informative?