Data Visualization Exercises in R: 25 Real-World Practice Problems
Twenty-five scenario-based problems covering chart types, distributions, customization, facets, annotations, and publication-ready polish in R. Every exercise ships with an expected result so you can verify your answer, and solutions stay hidden until you reveal them so you actually try first.
Section 1. Chart-type foundations (4 problems)
Exercise 1.1: Scatter plot of engine size against highway mileage
Task: Use the built-in mpg dataset (loaded with ggplot2) to build a scatter plot mapping displ on the x-axis and hwy on the y-axis with geom_point(). The plot should show all 234 observations with the default styling and no additional aesthetics. Save the result to ex_1_1.
Expected result:
#> A scatter plot with 234 black points.
#> x-axis 'displ' ranges roughly 1.5 to 7.0 (engine displacement in litres).
#> y-axis 'hwy' ranges 10 to 45 (highway mpg).
#> Strong negative trend: bigger engines deliver lower fuel economy.Difficulty: Beginner
Think of a plot as three pieces: the data, a mapping that says which column goes on which axis, and one layer that actually draws something.
Open the plot with ggplot(mpg, aes(x = displ, y = hwy)) and add geom_point() for the visible layer.
Click to reveal solution
Explanation: ggplot() opens the plot with the data and the aesthetic mapping; geom_point() adds the visible layer. Without color, size, or shape the default is a small filled circle. Memorize this two-line pattern: every ggplot starts as a data + mapping + geom triple, and every later customization adds a layer on top of this base.
Exercise 1.2: Color the scatter by drivetrain category
Task: Build on the previous scatter of displ against hwy from mpg, this time mapping drv to the color aesthetic inside aes() so each drivetrain (4, f, r) gets its own hue. ggplot2 should auto-generate the legend without any manual scale call. Save to ex_1_2.
Expected result:
#> Scatter with 234 points coloured by drivetrain.
#> Three colour groups: '4' (four-wheel), 'f' (front), 'r' (rear).
#> Legend on the right titled 'drv'.
#> Front-wheel cars (f) cluster at lower displacement and higher hwy mpg.Difficulty: Intermediate
The deciding question is whether the colour should respond to a data column or stay one fixed value for every point.
Add color = drv inside aes() so the hue is driven by the data, not inside geom_point().
Click to reveal solution
Explanation: Mapping color = drv inside aes() ties hue to the data; ggplot2 auto-picks a discrete palette and a legend. Contrast with geom_point(color = "blue") outside aes(), which paints every point the same colour regardless of data. The "inside or outside aes()" rule is the single most common stumbling block for new ggplot2 users.
Exercise 1.3: Line chart of US unemployment over time
Task: An ops engineer reviewing macro indicators wants a line chart of US unemployment from the economics dataset (loaded with ggplot2). Map date to x and unemploy to y with geom_line(). Add an informative y-axis label using scale_y_continuous(labels = scales::comma) so the thousands separator is shown. Save to ex_1_3.
Expected result:
#> A line chart spanning July 1967 to April 2015.
#> y-axis labelled with comma-separated thousands: e.g. 2,500, 5,000, 7,500, 10,000, 12,500, 15,000.
#> Major peaks visible during the early 1980s recession and the 2008 to 2010 downturn.Difficulty: Intermediate
A time series needs a layer that connects observations in date order, and large y-values read better with thousands separators.
Use geom_line() for the layer and scale_y_continuous(labels = scales::comma) for the formatted axis.
Click to reveal solution
Explanation: geom_line() connects observations in x-order, the right choice for a time series of monthly observations. The default y-axis prints values like 15000 which read awkwardly; scales::comma produces 15,000. The scales package is the standard companion to ggplot2 for axis formatters: comma, dollar, percent, label_log() all live there.
Exercise 1.4: Horizontal bar chart of diamonds by cut, sorted
Task: A jewellery retailer preparing an inventory dashboard wants a horizontal bar chart counting diamonds$cut rows, with bars sorted from longest to shortest. Use geom_bar(), then flip the chart with coord_flip() and reorder cut by its count using forcats::fct_infreq() so the longest bar sits at the top. Save to ex_1_4.
Expected result:
#> Horizontal bar chart with 5 bars.
#> Top to bottom: Ideal (~21,551), Premium (~13,791), Very Good (~12,082),
#> Good (~4,906), Fair (~1,610).
#> x-axis is 'count', y-axis is 'cut'.Difficulty: Intermediate
Let the geom do the counting for you, reorder the category by how often each level appears, then turn the chart on its side.
Map forcats::fct_infreq(cut) to x, add geom_bar(), and finish with coord_flip().
Click to reveal solution
Explanation: geom_bar() does the counting for you when no y is mapped, so a single discrete x is enough. fct_infreq() reorders the factor by descending frequency so the bars sort themselves; coord_flip() then rotates the whole plot. An alternative is geom_bar() + aes(y = forcats::fct_infreq(cut)) which avoids coord_flip() entirely in modern ggplot2.
Section 2. Distributions and density (4 problems)
Exercise 2.1: Histogram of highway mpg with fixed bin width
Task: Use mpg$hwy to draw a histogram with geom_histogram() and a fixed binwidth = 2. The plot should show the count of vehicles in each 2-mpg bin without any extra fills or facets. Save to ex_2_1.
Expected result:
#> Histogram of hwy with bin width 2 mpg.
#> x-axis 12 to 44, y-axis 0 to ~50.
#> Modal bin near 26 mpg; right tail extending to ~44 mpg with a few cars.Difficulty: Beginner
A count-per-interval chart needs you to decide how wide each interval should be rather than accept the default.
Map hwy to x and use geom_histogram(binwidth = 2).
Click to reveal solution
Explanation: geom_histogram() chooses 30 bins by default, which is rarely the right number. Setting binwidth to a unit you can interpret (here, 2 mpg) makes the histogram readable. Use binwidth for continuous axes; for already-discrete data prefer geom_bar(). A common alternative is bins = N if you want a fixed bin count.
Exercise 2.2: Density of ozone by month, faceted
Task: A climatologist preparing a seasonal report on the airquality dataset wants a density plot of Ozone faceted by Month. Convert Month to a factor first so each panel has a clean label, drop the NA Ozone rows with tidyr::drop_na(Ozone), then use geom_density(fill = "steelblue", alpha = 0.5) and facet_wrap(~ Month). Save to ex_2_2.
Expected result:
#> Five density panels, one per month (5 through 9).
#> Each panel shows the ozone distribution for that month.
#> July (7) and August (8) panels are visibly right-shifted versus May (5).
#> Fill is semi-transparent steelblue.Difficulty: Intermediate
Drop the missing values first, make the grouping variable categorical, then split the plot into one panel per group.
Chain tidyr::drop_na(Ozone), factor(Month), geom_density(fill = "steelblue", alpha = 0.5), and facet_wrap(~ Month).
Click to reveal solution
Explanation: geom_density() is a kernel-smoothed alternative to a histogram and works better when comparing multiple groups because the curves are continuous. alpha = 0.5 makes overlapping fills readable when you switch to overlay (no facet) later. Faceting on a factor is essential here: leaving Month numeric would give a continuous gradient and break the panels.
Exercise 2.3: Boxplot of mpg by cylinder count
Task: Use mtcars to draw a boxplot of mpg grouped by cyl, treating cyl as a factor so each cylinder count gets its own box. Map fill to the same factor so each box is coloured automatically by group. Use geom_boxplot(). Save the plot to ex_2_3.
Expected result:
#> Three boxes side by side for cyl 4, 6, 8.
#> Box for cyl 4 highest median (~26 mpg).
#> Box for cyl 8 lowest median (~15 mpg).
#> Each box filled with a distinct ggplot2 default colour.Difficulty: Intermediate
A discrete grouping stored as a number must be made categorical before it can split into separate boxes.
Wrap cyl in factor() for both the x mapping and the fill mapping, then add geom_boxplot().
Click to reveal solution
Explanation: Wrapping cyl in factor() is the canonical fix when a discrete categorical variable is stored as an integer; without it, ggplot2 treats cyl as continuous and draws one wide box that ignores the grouping. Mapping fill to the same factor is a one-line way to colour-code groups; if you mapped fill = cyl (continuous), you would get a colour gradient, not three distinct fills.
Exercise 2.4: Violin plus boxplot overlay of vitamin C response
Task: A pharmacology team analysing the ToothGrowth dataset wants both a violin (showing the full distribution shape) and a narrow inner boxplot for len by supp. Use geom_violin() first and then geom_boxplot(width = 0.1, fill = "white") so the box overlays the violin. Save the plot to ex_2_4.
Expected result:
#> Two side-by-side violins for supp 'OJ' and 'VC'.
#> Each violin has a narrow white boxplot drawn inside.
#> OJ violin centred slightly higher (median ~22) than VC (median ~17).
#> Violins show bimodal shape for VC reflecting dose levels.Difficulty: Intermediate
Two layers tell the story together, and the order you add them decides which one sits on top.
Add geom_violin() first, then geom_boxplot(width = 0.1, fill = "white") so the narrow box overlays the violin.
Click to reveal solution
Explanation: The violin shows the full kernel density on each side of the centre line; the inner boxplot gives you the conventional five-number summary for context. Layer order matters here: violin first, then boxplot, so the box sits on top. Setting width = 0.1 keeps the box thin so the violin shape stays visible behind it. Common alternative: replace the inner box with stat_summary(fun = median, geom = "point").
Section 3. Customization: labels, themes, scales (5 problems)
Exercise 3.1: Add title, subtitle, and axis labels
Task: Take a scatter of displ against hwy from mpg and add a labs() call with title = "Engine size vs highway mpg", subtitle = "234 model-year vehicles, 1999 and 2008", x = "Displacement (L)", and y = "Highway miles per gallon". Save the plot to ex_3_1.
Expected result:
#> Scatter plot with descriptive title above and subtitle below.
#> x-axis label reads 'Displacement (L)'.
#> y-axis label reads 'Highway miles per gallon'.
#> Plot shows the same negative trend as exercise 1.1.Difficulty: Beginner
Every piece of plot text - the heading, the sub-heading, the axis names - is set through one single call.
Add labs(title = ..., subtitle = ..., x = ..., y = ...) to the scatter.
Click to reveal solution
Explanation: labs() is the single entry point for every text label: title, subtitle, caption, tag, x, y, plus any legend titles by aesthetic name (color =, fill =). Use it instead of ggtitle(), xlab(), ylab() which still work but split your label setup across several calls. A complete labs() block makes the plot publication-ready in one line.
Exercise 3.2: Format y-axis with dollar labels on diamond prices
Task: A finance reporting analyst preparing a slide on diamond inventory wants a scatter of carat against price from diamonds with the y-axis formatted as US dollars. Use geom_point(alpha = 0.2) to handle the overplot, then add scale_y_continuous(labels = scales::dollar). Save to ex_3_2.
Expected result:
#> Scatter of 53,940 semi-transparent points (alpha 0.2 reveals density).
#> x-axis 'carat' 0 to 5.
#> y-axis tick labels formatted like '$5,000', '$10,000', '$15,000'.
#> Cone-shaped spread: prices fan upward at higher carat.Difficulty: Intermediate
Heavy overplotting calls for see-through points, and a money axis reads better with currency-formatted labels.
Use geom_point(alpha = 0.2) and scale_y_continuous(labels = scales::dollar).
Click to reveal solution
Explanation: alpha = 0.2 is the standard workaround when you have 50,000+ overlapping points: each individual point is barely visible but stacked points darken into the visible structure of the cloud. scales::dollar returns a formatter function that ggplot2 calls per tick; you can swap in percent, comma, or unit_format without changing the rest of the call.
Exercise 3.3: Apply theme_minimal and rotate x labels 45 degrees
Task: Take a count bar chart of diamonds$cut (use geom_bar()) and clean it up with theme_minimal(). Then rotate the x-axis text 45 degrees and right-justify it using theme(axis.text.x = element_text(angle = 45, hjust = 1)). Save the plot to ex_3_3.
Expected result:
#> Five vertical bars: Fair, Good, Very Good, Premium, Ideal.
#> Background is white with light grey gridlines (theme_minimal).
#> x-axis labels rotated 45 degrees and end-aligned at each tick.Difficulty: Intermediate
Apply the overall look first, then make the fine-grained adjustment to the tick text - order matters here.
Add theme_minimal() and then theme(axis.text.x = element_text(angle = 45, hjust = 1)).
Click to reveal solution
Explanation: theme_minimal() is one of about ten built-in themes; pick from theme_bw(), theme_classic(), theme_void() to match the look you want. Always apply the global theme BEFORE adding fine-grained theme() tweaks: if you call theme() first and theme_minimal() second, the minimal theme overwrites your tweaks. hjust = 1 right-anchors the rotated text so the tail of the label points at the tick.
Exercise 3.4: Continuous viridis color for clarity-graded scatter
Task: A jeweller exploring how clarity grade affects price-per-carat wants a scatter of carat against price from diamonds coloured by depth (a continuous variable). Map color = depth and apply the colour-blind-friendly continuous viridis palette via scale_color_viridis_c(). Save the plot to ex_3_4.
Expected result:
#> Scatter of 53,940 points coloured along the viridis (yellow to dark purple) gradient.
#> Legend on right titled 'depth' with a continuous colour bar.
#> Most points sit in the green-to-teal mid-range (depth around 61 to 62).Difficulty: Intermediate
A numeric colour variable needs a continuous palette, and there is a colour-blind-safe one built in.
Map color = depth and add scale_color_viridis_c().
Click to reveal solution
Explanation: Use _c for continuous data (gradient bar legend) and _d for discrete (one colour per category). Viridis is the default modern recommendation because it stays perceptually uniform under colour-blind simulation and prints well in grayscale. If your data is highly skewed, pair it with trans = "log10" to spread the colour resolution across the bulk of the range.
Exercise 3.5: Custom manual palette with three named colors
Task: Use mtcars to draw a scatter of wt against mpg with color = factor(cyl), then override the default ggplot2 palette with three hand-picked hex colours via scale_color_manual(values = c("4" = "#1b9e77", "6" = "#d95f02", "8" = "#7570b3")). Save the plot to ex_3_5.
Expected result:
#> Scatter of 32 points coloured by cyl.
#> 4-cylinder cars in teal (#1b9e77).
#> 6-cylinder cars in burnt orange (#d95f02).
#> 8-cylinder cars in muted purple (#7570b3).
#> Legend on right titled 'factor(cyl)'.Difficulty: Advanced
You can replace the automatic colours with your own, keyed so each category always maps to the same hue.
Add scale_color_manual(values = c("4" = "#1b9e77", "6" = "#d95f02", "8" = "#7570b3")).
Click to reveal solution
Explanation: Named-vector palettes give you full control and stay stable even if the data order changes. The three hex codes above come from the ColorBrewer "Dark2" qualitative palette, which is colour-blind friendly. If you don't care which level gets which colour, pass an unnamed vector: c("#1b9e77", "#d95f02", "#7570b3"). For larger sets, prefer scale_color_brewer(palette = "Dark2") so you don't enumerate by hand.
Section 4. Multi-plot composition and facets (4 problems)
Exercise 4.1: Facet a scatter by vehicle class
Task: Use mpg to build a scatter of displ against hwy, then split it into one panel per class using facet_wrap(~ class). Allow ggplot2 to pick the panel layout (it defaults to roughly square). Save the plot to ex_4_1.
Expected result:
#> Seven panels, one per vehicle class:
#> 2seater, compact, midsize, minivan, pickup, subcompact, suv.
#> Each panel shows the displ vs hwy scatter for that class only.
#> Layout is 3 columns, 3 rows with one empty cell.Difficulty: Beginner
Split the plot into one small panel per category of a single variable, wrapped into a grid.
Add facet_wrap(~ class) to the scatter.
Click to reveal solution
Explanation: facet_wrap() lays out one panel per level of the faceting variable and wraps them into rows. Control the grid with nrow = or ncol =; let scales vary with scales = "free", "free_x", or "free_y" when class-specific ranges drown out detail. Use facet_grid() instead when you have two faceting variables and want a true row-by-column matrix.
Exercise 4.2: Facet grid by drivetrain and year
Task: Continuing on mpg, build a scatter of displ against hwy then split with facet_grid(drv ~ year) so rows correspond to drivetrain (4, f, r) and columns to year (1999, 2008). This puts every drv-year cell side by side for easy cross-comparison. Save to ex_4_2.
Expected result:
#> Six-cell grid: 3 rows (drv: 4, f, r) by 2 columns (year: 1999, 2008).
#> Each cell shows the displ vs hwy scatter for that drv-year combo.
#> Strip labels: row strips on the right show '4', 'f', 'r'; column strips on top show '1999', '2008'.Difficulty: Intermediate
Splitting on two variables at once means a true rows-by-columns matrix of panels, not a wrapped strip.
Add facet_grid(drv ~ year) so drivetrain forms the rows and year the columns.
Click to reveal solution
Explanation: The formula rows ~ cols in facet_grid() is the easiest way to compare two factor variables jointly. Unlike facet_wrap(), the grid forces every row-column combination to appear even if it's empty, so visual gaps tell a story. Add scales = "free_y" when one row's range dwarfs another; otherwise the shared axis is helpful for cross-cell comparison.
Exercise 4.3: Side-by-side composition with patchwork
Task: A reporting analyst building a one-page dashboard wants three plots laid out side by side: a mpg$hwy histogram, a mpg$displ vs mpg$hwy scatter, and a count bar chart of mpg$class. Build each plot individually as p1, p2, p3, then combine them on a single row using the patchwork + operator. Save the combined object to ex_4_3.
Expected result:
#> Three panels arranged in a single row.
#> Left panel: histogram of hwy with default bins.
#> Centre panel: scatter of displ vs hwy.
#> Right panel: vertical bar chart counting classes.Difficulty: Advanced
Build each chart as its own independent object first, then join the finished objects into a single row.
Make p1, p2, p3 separately and combine them with p1 + p2 + p3.
Click to reveal solution
Explanation: patchwork overloads + to put plots side by side, / to stack them vertically, and | for an explicit row. Wrap groups in parentheses: (p1 + p2) / p3 puts two plots on top and one underneath. Use plot_layout(ncol = 2) to force a specific grid. patchwork respects each plot's own theme, scales, and legend, so build each piece independently and compose at the end.
Exercise 4.4: Free-scale facets on long-running housing series
Task: A housing analyst exploring txhousing wants a small-multiples line chart of median price over time for the five cities Austin, Dallas, Houston, San Antonio, and Fort Worth. Build a year + month/12 proxy for time, then facet by city with scales = "free_y" so each city uses its own y-range. Save to ex_4_4.
Expected result:
#> Five line panels, one per city.
#> x-axis 'time' spans roughly 2000 to 2015.
#> Each y-axis scales independently (free_y), so Houston's high prices don't flatten San Antonio.
#> Every line trends upward with a visible 2008 to 2010 dip.Difficulty: Advanced
When panels differ wildly in absolute level, each one should get its own y-range so the shape of the trend stays readable.
Filter to the five cities, build a year + month / 12 time column, and facet with facet_wrap(~ city, scales = "free_y").
Click to reveal solution
Explanation: Free scales (scales = "free_y") are the right answer whenever absolute levels differ wildly across panels but the shape of the trend is the question of interest. The trade-off is that a casual reader may misread the panels as similarly priced; always annotate the y-axis carefully in this case. For coordinated zoom across all panels, use scales = "fixed" (the default) and consider z-scoring or indexing each city to 100 at a baseline date.
Section 5. Annotations and statistical layers (4 problems)
Exercise 5.1: Add a linear smoother to a scatter plot
Task: Use mpg to draw a scatter of displ against hwy, then overlay a linear regression line with geom_smooth(method = "lm", se = TRUE) so the 95 percent confidence band is shown around the fit line. Save the plot to ex_5_1.
Expected result:
#> Scatter of 234 points (as in exercise 1.1).
#> A downward-sloping straight line overlaid (linear fit).
#> Grey ribbon around the line marking the 95% confidence interval.
#> Slope is clearly negative, roughly -3.5 mpg per litre of displacement.Difficulty: Intermediate
Add a second layer on top of the points that summarises the relationship with a fitted straight line and a confidence band.
Add geom_smooth(method = "lm", se = TRUE).
Click to reveal solution
Explanation: geom_smooth(method = "lm") fits an OLS line per group; without a color mapping, you get one global fit. Drop method = "lm" and geom_smooth() picks loess for small samples and gam for large ones; explicit is better than implicit. The grey ribbon is the confidence interval on the mean fit (turn off with se = FALSE), not the prediction interval for new points.
Exercise 5.2: Bar chart with count labels above each bar
Task: A sales analyst building a quarterly slide wants a count bar chart of mpg$class with the exact count printed above each bar. Use geom_bar() and add geom_text(stat = "count", aes(label = after_stat(count)), vjust = -0.5) so the label is anchored just above the bar top. Save the plot to ex_5_2.
Expected result:
#> Seven vertical bars (one per class) with the exact integer count drawn just above each bar.
#> Example labels above bars: '5' (2seater), '47' (compact), '41' (midsize), '11' (minivan),
#> '33' (pickup), '35' (subcompact), '62' (suv).Difficulty: Intermediate
The numbers you want printed are the bar heights, which are computed by the bar layer rather than present in the raw data.
Add geom_text(stat = "count", aes(label = after_stat(count)), vjust = -0.5).
Click to reveal solution
Explanation: geom_text() needs stat = "count" here because the y values (the bar heights) are computed by ggplot2 inside the bar layer, not present in the raw data. after_stat(count) exposes the computed stat to the label aesthetic; the older ..count.. syntax still works but is deprecated. vjust = -0.5 lifts the text above the bar; positive vjust pushes it inside the bar.
Exercise 5.3: Mean and standard-error bars with stat_summary
Task: A pharmacology analyst comparing OJ and VC supplements in ToothGrowth wants a clean summary chart: one point per supp at the group mean of len, with an error bar spanning plus and minus one standard error. Use stat_summary(fun = mean, geom = "point", size = 3) and stat_summary(fun.data = mean_se, geom = "errorbar", width = 0.2). Save to ex_5_3.
Expected result:
#> Two points (one for 'OJ', one for 'VC') on the x-axis.
#> Each point has a vertical error bar of width 0.2 centred on the mean.
#> OJ mean roughly 20.7, error bar half-width about 1.3.
#> VC mean roughly 17.0, error bar half-width about 1.4.Difficulty: Advanced
You can put summary statistics directly on the plot without computing them in a separate step beforehand.
Use two stat_summary() calls - one with fun = mean, geom = "point" and one with fun.data = mean_se, geom = "errorbar".
Click to reveal solution
Explanation: stat_summary() lets you put summary statistics on a plot without pre-computing them. fun = returns a single y for a point or line; fun.data = returns a data frame with y, ymin, ymax for error bars. Built-in summary helpers include mean_se, mean_cl_normal, mean_cl_boot, and median_hilow. Pair with geom_jitter(alpha = 0.3) underneath to show the raw points behind the summary.
Exercise 5.4: Reference lines and a focal annotation
Task: A trading-desk analyst reviewing the economics series wants the unemployment line annotated. Draw the line as before, add a horizontal reference line at yintercept = 10000 (10 million unemployed) with geom_hline(color = "red", linetype = "dashed"), and label one focal peak in 2009 with annotate("text", x = as.Date("2009-10-01"), y = 15500, label = "Great Recession peak"). Save to ex_5_4.
Expected result:
#> Line chart of unemployment vs date.
#> A dashed red horizontal line crosses the chart at y = 10,000.
#> A single text annotation 'Great Recession peak' sits above the 2009 spike.Difficulty: Advanced
A fixed horizontal marker and a single piece of placed text both annotate the chart without depending on the data rows.
Add geom_hline(yintercept = 10000, color = "red", linetype = "dashed") and annotate("text", x = ..., y = 15500, label = ...).
Click to reveal solution
Explanation: Prefer annotate() over geom_text() when you want exactly one label at a known position: geom_text() requires a data frame and emits one label per row, which is the right tool only for labelling many data points. geom_hline() and geom_vline() (and geom_abline()) are similar one-line annotators for reference lines that don't depend on the data.
Section 6. Publication-ready polish (4 problems)
Exercise 6.1: theme_classic with bold title and caption
Task: Take the mpg displ-vs-hwy scatter, apply theme_classic() for a Tufte-style spare look, then bold the title with theme(plot.title = element_text(face = "bold", size = 14)). Add a caption "Source: EPA fueleconomy.gov" via labs(caption = ...). Save to ex_6_1.
Expected result:
#> Spare black-and-white look (theme_classic): only x and y axis lines, no panel grid.
#> Title rendered in bold, size 14.
#> Caption 'Source: EPA fueleconomy.gov' anchored at bottom right.Difficulty: Intermediate
Pick the spare overall look first, then make one text element bold and add a source line at the bottom.
Use theme_classic(), theme(plot.title = element_text(face = "bold", size = 14)), and labs(caption = ...).
Click to reveal solution
Explanation: theme_classic() strips the panel grid and box for a Tufte-style minimal look that prints well in PDFs and slides. Always layer your custom theme() AFTER the named theme function so it overrides rather than gets overridden. Captions live at the bottom right by default; move them with theme(plot.caption = element_text(hjust = 0)) for left alignment.
Exercise 6.2: Save a 1200x800 PNG at 300 dpi
Task: Build any plot of your choice from mpg then save it to disk as mpg_report.png at 1200 pixels wide, 800 pixels tall, 300 dpi resolution, using ggsave(). Pass the plot object explicitly via the plot = argument so the call does not depend on a previous last_plot(). Save the plot object you built to ex_6_2.
Expected result:
#> A PNG file 'mpg_report.png' written to the working directory.
#> Image dimensions 1200 pixels by 800 pixels at 300 dpi.
#> ggsave prints a confirmation: 'Saving 4 x 2.67 in image' (approx, depends on width/height units).Difficulty: Advanced
Writing a plot to an image file is a single call where you state the file name, the dimensions, and the resolution.
Call ggsave(filename = "mpg_report.png", plot = ex_6_2, width = 1200, height = 800, units = "px", dpi = 300).
Click to reveal solution
Explanation: units = "px" was added in ggplot2 3.3 and is the cleanest way to specify pixel-exact output for web or slide deck use; the older units = "in" (inches) is still common in print workflows. ggsave() infers the file format from the extension; PNG, PDF, SVG, JPEG, and TIFF are all supported. For vector output that scales cleanly in slides, prefer .svg or .pdf over .png.
Exercise 6.3: Combined log-y, dollar labels, and custom palette
Task: A jewellery analytics team wants a publication-ready scatter of carat against price from diamonds, coloured by cut. Use a log10 y-axis with scale_y_log10(labels = scales::dollar), set the colour palette via scale_color_brewer(palette = "Set2"), and apply theme_minimal(). Save the plot to ex_6_3.
Expected result:
#> Scatter of 53,940 points coloured by cut (five categories).
#> y-axis on a log10 scale with dollar-formatted tick labels: '$500', '$1,000', '$2,000', '$5,000', '$10,000'.
#> Colour-blind safe Set2 palette (soft greens, oranges, pinks, blues).
#> Background white with light grey gridlines.Difficulty: Advanced
Three independent adjustments stack on the same plot - a compressed y-axis, money-formatted labels, and a chosen palette.
Add scale_y_log10(labels = scales::dollar), scale_color_brewer(palette = "Set2"), and theme_minimal().
Click to reveal solution
Explanation: Combining scale_y_log10() with a dollar formatter is the standard pattern for any right-skewed monetary scale: the log axis unstacks the bulk of the data while the formatter keeps tick labels human-readable. scale_color_brewer() picks from the ColorBrewer palettes; "Set2" is qualitative and colour-blind safe. For sequential or diverging palettes (ordered factors, residuals), see "Blues", "YlGnBu", or "RdBu".
Exercise 6.4: Flipped bar with value labels for a top-N report
Task: An HR analyst preparing a headcount report wants a horizontal bar chart of mpg class counts where each bar is labelled with its exact count at the bar tip. Use aes(y = forcats::fct_infreq(class)) for the y-axis, geom_bar(), geom_text(stat = "count", aes(label = after_stat(count)), hjust = -0.2), and stretch the x-axis with scale_x_continuous(expand = expansion(mult = c(0, 0.1))) so the labels are not clipped. Save to ex_6_4.
Expected result:
#> Horizontal bars, longest at top: suv, compact, midsize, subcompact, pickup, minivan, 2seater.
#> Each bar has its count printed just past the bar tip (e.g. '62' next to the suv bar).
#> x-axis padded on the right so labels are not cut off.Difficulty: Intermediate
Map the category to the vertical axis, sort it by frequency, label each bar tip, and pad the value axis so labels are not clipped.
Use aes(y = forcats::fct_infreq(class)), geom_text(stat = "count", aes(label = after_stat(count)), hjust = -0.2), and scale_x_continuous(expand = expansion(mult = c(0, 0.1))).
Click to reveal solution
Explanation: Mapping the discrete variable to y instead of x skips the older coord_flip() workaround and keeps the data flow clean. expansion(mult = c(0, 0.1)) adds zero padding on the left (so bars start at exactly zero) and 10 percent on the right (so the labels fit); the older expand_scale() is the same thing renamed. hjust = -0.2 positions the label just past the bar tip.
What to do next
You are ready for deeper, geom-specific drills. Three good next stops:
- ggplot2 Exercises in R for thirty more problems centred specifically on the grammar of graphics.
- ggplot2 Themes Exercises in R for full theme customization and brand-styling work.
- ggplot2 Facets Exercises in R for the small-multiples patterns that power any dashboard.
- EDA Exercises in R to put these chart-types to work on raw data discovery.
r-statistics.co · Verifiable credential · Public URL
This document certifies mastery of
Data Viz 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.
189 learners have earned this certificate