Data Visualization

How to Make Dynamic Figures

TipHow to render with/without answers

• With answers (default): quarto render exercises/data_viz.qmd
• Without answers (figures only): quarto render exercises/data_viz.qmd -P show_answers:false

ImportantWorking with agentic AI

An agent can help you explore an API, diagnose an error, or propose variants. It cannot make the visual argument for you. For every redesign prompt below, state the audience and decision the figure supports; keep an AI-use log (prompt, output used, and changes you made); and verify every displayed encoding against the data. Do not submit a plot you cannot explain layer by layer.

1) Distribution: Histogram + Density Overlay (Diamonds)

You are given ggplot2::diamonds (>50k rows). Create a histogram of price (restrict to $10,000 or less for visibility), overlay a scaled density curve so both share a comparable y-scale, and style axes and labels.

Requirements

• Filter to price <= 10000.
• Histogram with a binwidth of 250.
• Overlay a density where y = after_stat(count) * 250 so peaks align with histogram counts.
• Label axes and apply a dollar formatter to x-axis.
• Use theme_bw() and a concise title/subtitle.

library(ggplot2)

p1 <- diamonds |>
  filter(price <= 10000) |>
  ggplot(aes(price)) +
  geom_histogram(binwidth = 250, boundary = 0, closed = "left", color = "white") +
  geom_density(aes(y = after_stat(count) * 250), linewidth = 1, alpha = 0.15, fill = NA) +
  scale_x_continuous(labels = scales::dollar_format(accuracy = 1)) +
  labs(
    title = "Diamond Prices (≤ $10k)",
    subtitle = "Histogram with density overlay scaled to counts",
    x = "Price (USD)", y = "Count"
  ) +
  theme(legend.position = "none")

p1

Figure 1: Histogram and scaled density of diamond prices (≤ $10k).

---

1.1) Redesign for a decision

Choose a nontechnical buyer deciding whether a diamond is in a plausible price range. Redesign the plot for that decision, then write 100–150 words explaining what you changed, what trade-off it creates, and why the density overlay is or is not still useful. An agent may propose options, but you must justify the final choice.

2) Bivariate: Scatter with LOESS Trend + Facets (MPG)

Using ggplot2::mpg, visualize highway mileage (hwy) vs engine displacement (displ), colored by class. Add a LOESS smooth and facet by drive (drv).

Requirements

• geom_point(alpha = 0.6) and geom_smooth(se = FALSE, method = "loess").
• Axis limits: y from 10 to 45.
• Facet by drv with facet_wrap(~ drv, nrow = 1).
• Place legend at the bottom.

mpg |>
  ggplot(aes(displ, hwy, color = class)) +
  geom_point(alpha = 0.6) +
  geom_smooth(se = FALSE, method = "loess") +
  scale_y_continuous(limits = c(10, 45), breaks = seq(10, 45, 5)) +
  labs(
    title = "Fuel Economy by Engine Size",
    subtitle = "LOESS trends by vehicle class",
    x = "Engine displacement (L)", y = "Highway MPG"
  ) +
  facet_wrap(~ drv, nrow = 1) +
  theme(legend.position = "bottom")

Figure 2: Highway MPG vs. engine displacement, colored by class, faceted by drive (mpg).

---

2.1) Audit a misleading default

Create a deliberately less informative version of this plot (for example, by changing scales, overplotting, or grouping), then repair it. Include a short caption identifying the specific inference that the bad version could encourage.

3) Small Multiples: Facet Wrap by Class (MPG)

Produce small multiples of hwy vs displ but facet by class to compare patterns across vehicle classes.

Requirements

• Same base mapping as above; facet by class (ncol = 4).
• Use a consistent y-scale.
• Add a translucent linear fit per panel (method = "lm").

mpg |>
  ggplot(aes(displ, hwy)) +
  geom_point(alpha = 0.6, color = "grey30") +
  geom_smooth(method = "lm", se = FALSE, linewidth = 0.8, alpha = 0.9) +
  scale_y_continuous(limits = c(10, 45), breaks = seq(10, 45, 5)) +
  labs(
    title = "Fuel Economy vs Engine Size by Vehicle Class",
    x = "Engine displacement (L)", y = "Highway MPG"
  ) +
  facet_wrap(~ class, ncol = 4)

Figure 3: Small multiples of hwy vs displ by vehicle class (linear fits).

---

# 3.1) Make the above graph better

4) Time Series: Unemployment Rate with 12-Month Moving Average (economics)

Using ggplot2::economics, compute the unemployment rate as 100 * unemploy / pop. Plot the monthly rate and a 12-month moving average.

Requirements

• Create urate = 100 * unemploy / pop.
• Compute a 12-month moving average using stats::filter.
• Plot urate as a faint line and overlay the moving average as a thicker line.
• Limit the x-axis to 1970–2010 for focus.

library(stats)

economics |>
  mutate(
    urate = 100 * unemploy / pop,
    ma12 = as.numeric(stats::filter(urate, rep(1/12, 12), sides = 1))
  ) |>
  filter(date >= as.Date("1970-01-01"), date <= as.Date("2010-12-31")) |>
  ggplot(aes(date, urate)) +
  geom_line(alpha = 0.4) +
  geom_line(aes(y = ma12), linewidth = 1) +
  labs(
    title = "US Unemployment Rate",
    subtitle = "Monthly values with 12-month moving average",
    x = NULL, y = "Percent"
  )

Figure 4: US unemployment rate with 12-month moving average (1970–2010).

---

5) Distribution by Group: Violin + Box + Jitter (iris)

Using datasets::iris, compare Sepal.Length by Species with violins, embedded boxplots, and jittered points. Reorder species by median sepal length.

Requirements

• Reorder x-axis by median Sepal.Length.
• Use geom_violin(trim = FALSE, alpha = 0.3) + geom_boxplot(width = 0.1, outlier.shape = NA).
• Add geom_jitter(width = 0.1, alpha = 0.5, size = 0.8).
• Hide the fill legend and label clearly.

iris |>
  as_tibble() |>
  mutate(Species = fct_reorder(Species, Sepal.Length, .fun = median)) |>
  ggplot(aes(Species, Sepal.Length, fill = Species)) +
  geom_violin(trim = FALSE, alpha = 0.3) +
  geom_boxplot(width = 0.1, outlier.shape = NA) +
  geom_jitter(width = 0.1, alpha = 0.5, size = 0.8) +
  guides(fill = "none") +
  labs(
    title = "Sepal Length by Species",
    x = NULL, y = "Sepal length (cm)"
  )

Figure 5: Sepal length distributions by species (iris).

---

# 5.1) Make the above graph better

6) Correlation Heatmap with Clustering (mtcars)

Compute a Spearman correlation matrix for numeric variables in mtcars, cluster variables, and display a heatmap with correlation values.

Requirements

• Compute Spearman correlations on numeric columns.
• Order variables using hierarchical clustering on 1 - r.
• Use geom_tile() and scale_fill_gradient2() with limits c(-1, 1).
• Add rounded correlation labels.

num <- mtcars |> select(where(is.numeric))
R <- cor(num, method = "spearman", use = "pairwise.complete.obs")
ord <- hclust(as.dist(1 - R))$order
lvl <- colnames(R)[ord]

corr_df <- as.data.frame(R) |>
  rownames_to_column("var1") |>
  pivot_longer(-var1, names_to = "var2", values_to = "r") |>
  mutate(
    var1 = factor(var1, levels = lvl),
    var2 = factor(var2, levels = lvl)
  )

ggplot(corr_df, aes(var2, var1, fill = r)) +
  geom_tile() +
  geom_text(aes(label = sprintf("%.2f", r)), size = 3) +
  scale_fill_gradient2(limits = c(-1, 1)) +
  coord_fixed() +
  labs(x = NULL, y = NULL, fill = "rho",
       title = "Spearman Correlation (mtcars)") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

Figure 6: Spearman correlation heatmap with hierarchical clustering (mtcars).

---

# 6.1) Make the above graph better

7) Tile Heatmap: Housing Prices by Month × Year (txhousing)

Using ggplot2::txhousing, create a month × year heatmap of median home prices for Austin.

Requirements

• Filter city to "Austin".
• Limit years to 2000–2015.
• Y-axis should be month (Jan–Dec).
• Fill tiles by median (median sales price).
• Add a readable color scale and labels.

# subset and prepare month labels
au <- txhousing |>
  filter(city == "Austin", year >= 2000, year <= 2015) |>
  mutate(Month = factor(month, levels = 1:12, labels = month.abb))

ggplot(au, aes(year, Month, fill = median)) +
  geom_tile() +
  scale_fill_gradient(labels = scales::dollar, name = "Median price") +
  scale_x_continuous(breaks = seq(2000, 2015, by = 3)) +
  labs(
    title = "Austin Housing — Median Price by Month and Year",
    x = NULL, y = NULL
  ) +
  theme(legend.position = "right")

Figure 7: Median home price (Austin): month × year heatmap (txhousing).

# 7.1) Make the above graph better

8) Mini data-story: critique, build, and verify

Choose one dataset used above or a small, public dataset relevant to your interests. Pose one answerable descriptive question for a named audience. Create a two-figure data story: one overview and one supporting or qualifying view. Each figure needs an informative title, a caption that states the takeaway, and accessible color/label choices.

Use an AI agent for at least one bounded task (such as proposing a chart type, checking a ggplot2 API detail, or reviewing labels). Submit the prompt and a brief verification note. Your final write-up (250–400 words) must explain why each encoding is appropriate, name one limitation, and identify one change you rejected after inspecting the data. Put the analysis in a reproducible .qmd or .R file.

Appendix: Session Info (hidden)