ggplot2 Facet Exercises in R: 15 Real-World Practice Problems
Fifteen hands-on exercises that drill every facet skill in ggplot2: facet_wrap() layout control, facet_grid() for two-variable matrices, free vs fixed scales, custom strip labels, margins, and the background-data overlay trick. Each problem ships with a Task, an expected output, and a hidden step-by-step solution.
Run the setup block once, then attempt each exercise in the Your-turn box before revealing the solution. Sections build from basic facet_wrap() through advanced strip-label customization and the background-overlay pattern used in published dashboards.
Section 1. facet_wrap() basics (3 problems)
Exercise 1.1: Split mpg by vehicle class with facet_wrap
Task: A car-review editor wants one mini scatterplot of highway MPG versus engine displacement for every vehicle class in the mpg dataset, so readers can compare classes side by side. Use facet_wrap() to split the plot by the class column and save the ggplot object to ex_1_1.
Expected result:
#> ggplot output saved as ex_1_1.
#> 7 panels (one per vehicle class: 2seater, compact, midsize, minivan, pickup, subcompact, suv).
#> Default wrap layout puts panels in 3 columns by 3 rows (last row partial).
#> Within each panel: hwy on y, displ on x, blue points.Difficulty: Beginner
Splitting one chart into a small panel per category is exactly what faceting is for - decide which column defines the panels.
Add a facet layer keyed on the class column to a geom_point() plot of hwy versus displ.
Click to reveal solution
Explanation: facet_wrap(~ class) takes one categorical variable and lays out one panel per level. Because the formula has nothing on the left, ggplot picks a near-square grid (ncol defaults to a value close to sqrt(n_levels)). The same x and y scale is reused across panels, which makes cross-panel comparison fair. If groups vary wildly in range you would use scales = "free_y" instead.
Exercise 1.2: Force a single-row layout with ncol
Task: The editorial team wants the seven mpg class panels arranged in a single horizontal strip across the page for a newspaper print column. Rebuild the plot from Exercise 1.1 and force facet_wrap() to lay all panels in one row using the ncol argument. Save the result to ex_1_2.
Expected result:
#> ggplot output saved as ex_1_2.
#> 7 panels in a single row (1 row by 7 columns).
#> Panel strips read: 2seater | compact | midsize | minivan | pickup | subcompact | suv.
#> Each panel is narrow; x-axis tick labels may overlap at smaller widths.Difficulty: Beginner
The default near-square panel arrangement can be overridden so every panel lands in one horizontal strip.
Pass ncol = 7 to facet_wrap() so all seven class panels sit in a single row.
Click to reveal solution
Explanation: Passing ncol = 7 overrides the auto-layout and forces a single row. The mirror argument is nrow. Use whichever fits the rendering width: ncol for wide-aspect outputs (print, dashboards), nrow for tall columns (mobile, reports). When both are supplied, nrow wins because facet_wrap fills column-first. Tick density gets crowded at this width, so consider rotating labels with theme(axis.text.x = element_text(angle = 45, hjust = 1)).
Exercise 1.3: Facet diamonds price by cut quality
Task: A jeweller analyzing inventory wants one histogram of price per cut quality from the diamonds dataset, arranged in a 2-by-3 grid. Build a geom_histogram() plot and facet by cut, forcing exactly 3 columns. Save the ggplot to ex_1_3.
Expected result:
#> ggplot output saved as ex_1_3.
#> 6 panels in a 2-row by 3-col grid (5 cut levels: Fair, Good, Very Good, Premium, Ideal; one cell blank).
#> Each panel: price histogram, right-skewed, bulk under $5000.
#> "Fair" panel shows the widest spread; "Ideal" the most peaked at low price.Difficulty: Intermediate
One distribution chart per category can be flowed into a grid whose column count you fix yourself.
Build a geom_histogram() of price and facet by cut with ncol = 3.
Click to reveal solution
Explanation: facet_wrap accepts ordinal factors (cut is ordered) and preserves the factor order in panel arrangement. Because 5 panels do not fit a 3-column grid evenly, the bottom-right cell stays blank. To fill that gap you would either set ncol = 5 (single row) or use as.character(cut) and trim a level. label_dollar() from the scales package formats the x-axis without manually typing $ prefixes.
Section 2. facet_grid() for two-variable matrices (3 problems)
Exercise 2.1: Cross drive type and cylinder count in a grid
Task: A reliability analyst wants a strict matrix of MPG distributions, where rows correspond to drive type (drv) and columns to cylinder count (cyl) in the mpg dataset. Build a geom_point() plot of hwy versus displ and use facet_grid(drv ~ cyl). Save to ex_2_1.
Expected result:
#> ggplot output saved as ex_2_1.
#> 3 rows (drv: 4, f, r) by 4 cols (cyl: 4, 5, 6, 8) = 12 panels.
#> Some panels empty (e.g. 4wd + 5cyl, rwd + 4cyl don't exist in the data).
#> Each populated panel: hwy vs displ scatter.Difficulty: Intermediate
When two variables both matter, arrange them as a rigid rows-by-columns matrix rather than a flowing wrap.
Use facet_grid(drv ~ cyl) with the row variable before the tilde and the column variable after.
Click to reveal solution
Explanation: facet_grid(row_var ~ col_var) lays variables in a rigid matrix, which is the right choice when both dimensions are interesting and you want a quick crosstab. Unlike facet_wrap, empty combinations are still rendered as blank cells, which actually communicates that the combination is rare or impossible. To suppress empty cells use facet_grid(... , drop = TRUE); to free the row or column extents use scales = "free_x" or "free_y".
Exercise 2.2: Facet rows only with the dot placeholder
Task: A reviewer wants a stack of three panels (one per drive type) showing the highway MPG distribution as boxplots, all sharing the same x-axis. Use facet_grid() with the dot placeholder so drv becomes the row variable and there is no column variable. Save the plot to ex_2_2.
Expected result:
#> ggplot output saved as ex_2_2.
#> 3 rows, 1 column (drv: 4, f, r from top to bottom).
#> Each row: horizontal boxplot of hwy.
#> Strip labels on the right edge: 4, f, r.Difficulty: Intermediate
A matrix layout can split on a single axis only, with the other axis carrying no faceting variable at all.
Use facet_grid(drv ~ .) where the dot placeholder stands in for the missing column variable.
Click to reveal solution
Explanation: The . placeholder tells facet_grid that there is no faceting variable for that axis. drv ~ . means rows by drv, no column splits. The mirror is . ~ drv for columns only. This is cleaner than facet_wrap(~ drv, ncol = 1) when you want strip labels on the right edge rather than at the top of each panel, which is the facet_grid default.
Exercise 2.3: Cut by color crosstab on diamonds
Task: A jeweller wants a price-versus-carat scatter for every cell of a cut (5 levels) by color (7 levels) grid on the diamonds dataset, with rows as cut and columns as color. Use facet_grid() and reduce point size to handle the large rendering. Save the plot to ex_2_3.
Expected result:
#> ggplot output saved as ex_2_3.
#> 5 rows (cut: Fair, Good, Very Good, Premium, Ideal) by 7 cols (color: D...J) = 35 panels.
#> Each cell: price vs carat scatter with tiny points.
#> Pattern: tighter linear cloud in "Ideal + D"; loose spread in "Fair + J".Difficulty: Intermediate
A 35-cell matrix only stays readable if the points and text shrink to match the crowded layout.
Use facet_grid(cut ~ color) and tune geom_point() with a tiny size, low alpha, plus a reduced base_size.
Click to reveal solution
Explanation: When the grid is large (here 35 panels), small visual choices matter: tiny size, low alpha, smaller base_size so strip labels do not collide. facet_grid keeps all panels on the same x and y scale by default, which is exactly right for a crosstab where the comparison is "how does this relationship shift as I move across the grid?". If panels share little in common (say, different metrics on y), reach for scales = "free_y".
Section 3. Free vs fixed scales (3 problems)
Exercise 3.1: Let each MPG class have its own y-axis range
Task: A reviewer noticed that 2-seater highway MPG sits in a tight band while pickup MPG covers a wide low range, so the shared y-axis from Exercise 1.1 wastes vertical space in most panels. Rebuild the mpg facet plot and free the y-axis per panel with scales = "free_y". Save to ex_3_1.
Expected result:
#> ggplot output saved as ex_3_1.
#> 7 panels, each with an independent y-axis range.
#> 2seater panel y-range narrow (around 20-30 mpg).
#> Pickup panel y-range wider (around 15-25 mpg).
#> X-axis (displacement) still shared across all panels.Difficulty: Intermediate
When panels differ widely in range, a shared axis wastes vertical space - each panel can hold its own range instead.
Add scales = "free_y" to the facet_wrap(~ class) layer.
Click to reveal solution
Explanation: scales = "free_y" releases the y-axis per panel while keeping x shared. The four options are "fixed" (default), "free_x", "free_y", and "free" (both). Free scales make within-panel patterns easier to read but they kill cross-panel comparison: a slope of 5 in one panel may look identical to a slope of 50 elsewhere. Use free scales when each panel tells its own story; use fixed when readers should compare magnitudes across panels.
Exercise 3.2: Free both axes on a multi-metric panel built from inline data
Task: A monitoring dashboard tracks three unrelated server metrics (CPU percent, memory GB, request count) over the same 24 hours, but the metric ranges differ by orders of magnitude. Build the inline tibble below, pivot to long form, then facet by metric with scales = "free" so each metric reads on its own axes. Save the plot to ex_3_2.
Expected result:
#> ggplot output saved as ex_3_2.
#> 3 panels (one per metric), each with its own y-axis scale.
#> cpu_pct panel: y-axis roughly 0-60.
#> mem_gb panel: y-axis roughly 7.5-13.5.
#> req_cnt panel: y-axis roughly 0-1500.
#> All panels share the x-axis (hour 0-23).Difficulty: Advanced
Metrics in incompatible units need fully independent axes per panel, and the wide table must first be reshaped so one column names the panel.
Reshape with pivot_longer(), then facet the result with scales = "free" inside facet_wrap().
Click to reveal solution
Explanation: scales = "free" frees both axes per panel, which is the standard pattern for stacked time-series dashboards where metrics share the time axis but live in incompatible units. Pivoting to long form is the canonical setup: ggplot wants one row per (panel, observation). The alternative is patchwork::wrap_plots() of three separate ggplot objects, which gives finer control over individual panel themes but loses the shared time axis that facet_wrap provides for free.
Exercise 3.3: Proportional column widths with space="free_x"
Task: A product manager wants a bar chart of average price per clarity level on diamonds, faceted by cut in columns, where each column's width is proportional to how many clarity levels appear in that cut. Use facet_grid with both scales = "free_x" and space = "free_x". Save the plot to ex_3_3.
Expected result:
#> ggplot output saved as ex_3_3.
#> 1 row, 5 columns (one per cut: Fair...Ideal).
#> Each column shows bar chart of mean price by clarity (8 clarity levels possible).
#> Column widths vary: panels with fewer distinct clarity levels are narrower.
#> X-axis (clarity) labels appear only inside each panel, no shared scale.Difficulty: Advanced
Freeing the x range alone leaves panels equal width - the panel width itself can also be made proportional to the data it holds.
Combine scales = "free_x" with space = "free_x" inside facet_grid(~ cut).
Click to reveal solution
Explanation: scales = "free_x" alone would make each panel the same width but with different tick sets, leaving awkward whitespace. Adding space = "free_x" rescales the panels themselves so each bar gets equal horizontal space across the whole plot. This is the trick for nested-bar layouts (e.g. tasks grouped by sprint, where some sprints have more tasks): readers see honest visual proportion rather than padded panels.
Section 4. Custom strip labels (3 problems)
Exercise 4.1: Replace cryptic class codes with full names via as_labeller
Task: The mpg class strip labels read "2seater", "compact", "midsize", "minivan", "pickup", "subcompact", "suv", which are too terse for a public report. Build a named vector that maps each code to a polished label ("Two-Seater", "Compact Car", etc.), then pass it to as_labeller() inside facet_wrap(). Save the plot to ex_4_1.
Expected result:
#> ggplot output saved as ex_4_1.
#> 7 panels with cleaner strip labels.
#> Strips read: "Two Seater", "Compact Car", "Midsize Sedan", "Minivan", "Pickup Truck", "Subcompact", "SUV".
#> Plot content unchanged from Exercise 1.1.Difficulty: Intermediate
Panel strip text can be remapped to friendlier labels without changing the underlying data values.
Build a named vector mapping each code to a polished label and pass it via labeller = as_labeller(...) to facet_wrap().
Click to reveal solution
Explanation: as_labeller() accepts a named character vector where names match the data values and values are the desired labels. Anything missing from the vector falls back to the raw value, which is forgiving when categories are added later. For programmatic transforms (e.g. uppercase, title case) pass a function instead: labeller = as_labeller(toupper). For complex two-line strip text with variable name plus value use labeller = label_both (see Exercise 4.2).
Exercise 4.2: Show variable name and value with label_both
Task: A teaching slide needs strip labels in facet_grid(drv ~ cyl) to read "drv: 4" and "cyl: 6" rather than just "4" or "6", so students can tell which strip belongs to which variable. Rebuild the grid from Exercise 2.1 and pass labeller = label_both. Save to ex_4_2.
Expected result:
#> ggplot output saved as ex_4_2.
#> 12-cell grid as in Exercise 2.1.
#> Row strips read: "drv: 4", "drv: f", "drv: r".
#> Column strips read: "cyl: 4", "cyl: 5", "cyl: 6", "cyl: 8".Difficulty: Beginner
Strip text can include the variable name alongside its value so each panel explains which split it belongs to.
Pass labeller = label_both to facet_grid(drv ~ cyl).
Click to reveal solution
Explanation: label_both is a ready-made labeller function that prefixes each strip with its variable name. It is the cheapest way to make a multi-variable grid self-documenting, especially for audiences who will not see the axis legend. Cousins are label_value (default, value only), label_parsed (parses strings as math expressions, see Exercise 4.3), and label_bquote (substitutes values into a math expression template).
Exercise 4.3: Render Greek letters in strip labels with label_parsed
Task: A statistics handout needs facet strips to render as Greek symbols ("alpha = 0.05", "alpha = 0.10") with alpha displayed as the Greek letter. Build the inline tibble below, then facet by alpha_level using labeller = label_parsed so plotmath syntax is rendered. Save to ex_4_3.
Expected result:
#> ggplot output saved as ex_4_3.
#> 2 panels side by side, each plotting power vs n.
#> Strip labels render with the Greek letter alpha, e.g. "alpha = 0.05" appears as the symbol.
#> Power curves both monotonic increasing, the 0.10 panel sits below the 0.05 panel.Difficulty: Advanced
Strip text can be run through R's math-rendering parser so labels display actual symbols instead of raw strings.
Facet on alpha_level with labeller = label_parsed, using plotmath strings such as "alpha == 0.05".
Click to reveal solution
Explanation: label_parsed runs each strip label through R's plotmath parser, so strings like "alpha == 0.05" render as the Greek alpha followed by an equals sign and the value. The double == is plotmath syntax for an equals sign symbol (single = is reserved for assignment in expressions). When labels are dynamic, build them with sprintf or paste: paste0("alpha == ", round(a, 2)).
Section 5. Advanced and combination workflows (3 problems)
Exercise 5.1: Add a "Total" margin panel with margins = TRUE
Task: A pivot-table-style report needs a facet_grid() plot of mpg MPG by drv (rows) and cyl (columns) that also includes summary panels showing the aggregate across all rows and all columns (like a spreadsheet "Total" row and column). Set margins = TRUE on facet_grid. Save the plot to ex_5_1.
Expected result:
#> ggplot output saved as ex_5_1.
#> Original 12-cell grid plus extra row and column for "(all)" totals.
#> Bottom row strip reads "(all)" and aggregates all cylinder values.
#> Rightmost column strip reads "(all)" and aggregates all drive types.
#> Bottom-right corner cell shows the full data combined.Difficulty: Advanced
A grid can gain spreadsheet-style total panels that aggregate across every row and every column.
Set margins = TRUE on the facet_grid(drv ~ cyl) layer.
Click to reveal solution
Explanation: margins = TRUE duplicates the data into extra synthetic categories named (all) for each faceting variable, giving you the "Total" panels seen in pivot tables. To add margins only for one variable, pass the variable name as a string: margins = "drv". The trade-off is that data is duplicated, which inflates rendering time on large datasets and can mislead naive readers into thinking the totals are independent samples.
Exercise 5.2: Overlay all-data context behind each facet panel
Task: A reviewer wants every mpg class panel to show its own points in steel-blue plus the full mpg dataset in grey as background context, so readers see where the class sits within the whole cloud. Build a plot that calls geom_point() twice: once with the full mpg minus the facet column, once with the regular data. Save to ex_5_2.
Expected result:
#> ggplot output saved as ex_5_2.
#> 7 panels (one per class).
#> Each panel: light-grey backdrop of all 234 mpg points + colored foreground for that class.
#> The colored cluster sits within the grey cloud, making the class's region obvious.Difficulty: Advanced
To show each panel against the whole dataset, draw the full data once with the facet column removed so it repeats identically in every panel.
Call geom_point() twice - the first with data = transform(mpg, class = NULL) for the grey backdrop, the second with the normal data on top.
Click to reveal solution
Explanation: Setting class = NULL on the background data strips the facet variable so the first geom_point is drawn identically in every panel, producing the grey context cloud. The second geom_point keeps class and is therefore subset by facet_wrap to only that panel's points. This trick is one of the highest-impact reads in dashboard design: instead of asking readers to mentally overlay panels, you show the global cloud once and highlight the focal subset.
Exercise 5.3: Use facet_wrap with strip.position to mimic a small-multiple bar chart
Task: A finance analyst wants a quarterly revenue bar chart split into one mini-panel per region, with the region label shown along the bottom of each panel (not the default top), to match a published report style. Build the inline tibble below and facet by region with strip.position = "bottom". Save the plot to ex_5_3.
Expected result:
#> ggplot output saved as ex_5_3.
#> 4 panels in a single row (one per region).
#> Each panel: bar chart of rev_m by quarter (Q1...Q4).
#> Strip labels (region names) appear under each panel, not above.
#> Bars colored by quarter for visual variety.Difficulty: Intermediate
The panel label strip does not have to sit on top - it can be relocated to a different edge of each panel.
Pass strip.position = "bottom" to facet_wrap() and pair it with theme(strip.placement = "outside").
Click to reveal solution
Explanation: strip.position = "bottom" flips the default top strip to the bottom, which matches how many published bar-chart series are typeset (region name beneath the bars, like a caption). Pair it with theme(strip.placement = "outside") so the strip sits below the axis text rather than between the axis and the bars. The four valid positions are "top" (default), "bottom", "left", and "right". Use "left" or "right" for stacked rows of horizontal bar charts.
What to do next
- Review the parent guide: ggplot2 Facets covers theory and the function reference.
- Practice color and theme work: ggplot2 Theme Exercises.
- Build the underlying charts: ggplot2 Bar Chart Exercises and ggplot2 Scatter Plot Exercises.
- Apply faceting to real EDA: Exploratory Data Analysis Exercises in R.
r-statistics.co · Verifiable credential · Public URL
This document certifies mastery of
ggplot2 Facets (15 problems) Mastery
Every certificate has a public verification URL that proves the holder passed the assessment. Anyone with the link can confirm the recipient and date.
56 learners have earned this certificate