ggplot2 Aesthetics Exercises: 18 aes() Practice Problems with Solutions

Explanation: colour = class lives inside aes() because it is a data-driven mapping: ggplot2 assigns a different colour to each level of class. Putting colour = "blue" outside aes() would paint every point the same blue and silently break the grouping. The default discrete palette is hue-based; you can swap it later with scale_colour_brewer() or scale_colour_viridis_d().

Explanation: For bars, fill paints the interior while colour would only outline the bar edges, which is rarely what you want. geom_bar() defaults to stat = "count", so you only supply the x variable; ggplot2 tallies rows. To switch from stacked to side-by-side, add position = "dodge". To compare proportions across drivetrains rather than absolute counts, use position = "fill".

Explanation: This exercise mixes a data-driven mapping (fill = cut inside aes()) with a constant aesthetic (colour = "white" outside aes() inside geom_histogram()). That is the canonical way to apply a fixed visual style on top of a mapping. Putting colour = "white" inside aes() would attempt to treat the literal string "white" as a one-level factor and produce a single red line plus a misleading legend entry.

Explanation: size inside aes() produces a continuous size scale when the variable is numeric; ggplot2 maps the range of cyl to a default radius range of 1 to 6 mm. The variable cyl is technically discrete (only 4 values), so you may prefer as.factor(cyl) if you want a discrete legend with one entry per cylinder count. To set every point to the same size, write geom_point(size = 3) outside aes().

Explanation: Shape is a discrete-only aesthetic; ggplot2 will warn if you try to map a continuous variable. The default palette cycles through 6 shapes, after which extra levels are dropped with a warning. For more than 6 categories you must supply your own palette via scale_shape_manual(values = c(...)). Shape combined with colour is a common accessibility pattern (colour-blind users distinguish shapes even when colours blur).

Explanation: Alpha is the transparency channel: 0 is invisible, 1 is fully opaque. Setting alpha outside aes() is the standard fix for overplotting, because 10 overlapping 10%-opacity points sum to a fully opaque pixel and progressively darker shades mark density. If you map alpha inside aes() to a variable instead, you get a transparency gradient with a legend, which is rarely as readable as geom_hex() or geom_density_2d() for raw density.

Explanation: The single most common ggplot2 mistake. When colour = "steelblue" sits inside aes(), ggplot2 treats the literal string "steelblue" as a single-level factor, maps it through the default hue scale (which starts at salmon), and creates a one-entry legend. To set a fixed visual property, the argument goes OUTSIDE aes(), as a literal argument to geom_point(). Mnemonic: inside aes() is data-driven, outside aes() is a constant.

Explanation: Three aesthetics, two of them constant. fill = class is the only data-driven mapping and is the only one inside aes(). colour = "black" and width = 0.7 are both fixed values handed to geom_bar() directly. A common pitfall: writing aes(colour = "black") would draw red outlines and create a useless legend entry labelled "black". Mapping and setting always coexist in production charts; you must train the eye to spot which is which.

Explanation: This is an advanced pattern called "identity scale": you compute the colour string per-row inside aes(), then scale_colour_identity() tells ggplot2 to interpret those strings as literal R colours rather than running them through a palette. Without scale_colour_identity(), the levels "red" and "grey50" would be mapped through the default hue palette to salmon and teal, which is the opposite of what you want. To add a legend back, use scale_colour_identity(guide = "legend") with breaks and labels.

Explanation: ggplot2 automatically generates one legend per data-driven aesthetic, so mapping both colour and size yields two legends. Beware of cognitive load: charts with more than 3 mapped aesthetics quickly become unreadable. A clean alternative is to facet on one variable and keep only colour or size as a mapping. The constant alpha = 0.7 reduces overplotting without itself creating a legend.

Explanation: Both aesthetic arguments are constants here (no aes() wrappers), so no legend appears. linetype accepts named strings ("solid", "dashed", "dotted", "dotdash", "longdash", "twodash") or integers 0 to 6. geom_hline() inherits the parent ggplot but only consumes yintercept; the x mapping is ignored. To make the reference legend-visible, you would map a constant string inside aes() and customise the scale, which is rarely worth it for a single reference line.

Explanation: When two aesthetics map to the same variable, ggplot2 merges their legends into a single combined guide. This is a useful accessibility pattern: a reader who cannot distinguish red from green can still tell the lines apart by dash pattern. The group = region is technically redundant when colour already groups, but it is good practice to make grouping explicit, especially when adding geom_smooth() or geom_ribbon() layers that do not auto-group.

Explanation: scale_colour_viridis_c() is the continuous variant; the _d() suffix is for discrete data. Viridis is the standard accessible palette because it is perceptually uniform (equal data steps look like equal colour steps) and remains distinguishable for the most common forms of colour-blindness. The continuous mapping automatically produces a vertical colour bar guide rather than a discrete keyed legend.

Explanation: scale_fill_manual(values = ...) takes a NAMED vector where names match factor levels. The trick here is to build that named vector by starting from scales::hue_pal() (ggplot2's default discrete palette) and overwriting three entries with brand colours. Without naming the vector, the order must match the factor levels exactly. This pattern is common in client reports where corporate brand colours are non-negotiable for a subset of categories.

Explanation: range = c(min_mm, max_mm) controls the radius span in millimetres; the default 1 to 6 is often too subtle for presentations. breaks cherry-picks the values shown in the legend without changing the data mapping itself. For area-proportional bubbles (where data is mapped to area, not radius), use scale_size_area() instead, which is the better choice when bubble area is meant to represent a quantity (population, revenue).

Explanation: The grey-out pattern is a standard storytelling device: it suppresses irrelevant detail and lets one category dominate visually. The ifelse() inside aes() computes a per-row colour string, and scale_colour_identity() interprets those literal strings as R colours. For multi-category highlights, swap ifelse() for dplyr::case_when() or pre-compute a highlight column before plotting.

Explanation: Factor level order controls three things at once: legend order (top to bottom), stack order (bottom to top of the bar), and dodge order (left to right). Reordering with factor(x, levels = ...) or forcats::fct_relevel() is therefore the lever for visual hierarchy. A common gotcha: ggplot2 stacks the FIRST factor level at the BOTTOM of the bar; reversing the legend pushes "Ideal" to the bottom of each bar, which is usually the position the eye scans to first.

Explanation: This problem ties every concept in the hub together: one data-driven mapping (colour = cut), two constants outside aes() (alpha, size), a custom scale (scale_colour_viridis_d() for an accessible discrete palette), and a theme override. The order of layers matters for legibility, but the order of scale and theme additions does not. To override the legend dot opacity (so legend swatches look solid even though points are 30% alpha), append guides(colour = guide_legend(override.aes = list(alpha = 1, size = 3))).