Layering Multiple Geoms

We have seen that a plot is a stack of layers. Now we make that practical: combining several geoms into a single, information-rich figure, and controlling exactly which data and mappings each layer uses.

Stacking geoms that share a mapping

The simplest multi-layer plot adds geoms that all read the base mappings. Points + a smoother is the canonical example:

Both layers inherit x = displ, y = hwy from ggplot(). The points show the raw data; the smoother summarizes the trend. Together they say more than either alone — here are the data, and here is the pattern in them.

Inheritance: layers borrow from the base by default

A key rule: layers inherit the data and mappings declared in ggplot() unless you override them. This diagram shows the flow:

So a color = drv mapping in ggplot() reaches every layer:

Scoping a mapping to one layer

What if you want colored points but a single overall trend line, not one per group? Move the color mapping out of ggplot() and into the point layer only:

The points are colored by drivetrain; the smoother — which never saw color = drv — fits one black line to all the data. Where you place a mapping decides which layers it affects. This is the lever behind most sophisticated ggplot figures.

Overriding inherited data for one layer

A layer can also swap in different data. This is how you annotate or highlight. Here we label only the most fuel-efficient cars:

Three layers: all cars in grey, the standout cars re-emphasized in red, and text labels — the red layer and the text layer both override the base data with best, while still inheriting the x/y mappings.

Order still matters

Because layers composite in order, put the marks you want on top last. In the highlight example, drawing the grey points first and the red points second ensures the red ones are never hidden behind grey ones.

QuestionSelect one

You write ggplot(mpg, aes(displ, hwy, color = drv)) + geom_point() + geom_smooth(method = "lm"). How many trend lines appear, and why?

One line, because geom_smooth ignores color.

Three lines — one per drivetrain — because color = drv is declared in ggplot() and therefore inherited by the smoother, which groups by it.

No lines, because geom_smooth needs its own aes().

Two lines, one for points and one for smooth.

QuestionSelect one

You want points colored by drv but only one overall trend line. What is the correct change?

Add color = "black" inside ggplot(aes(...)).

Move the color mapping into the point layer: geom_point(aes(color = drv)), and leave color out of ggplot() so the smoother fits all data together.

Delete the geom_smooth layer.

Use inherit.aes = FALSE on geom_point.

Key takeaways

• Layers inherit the data and mappings from ggplot() unless overridden.
• A mapping in ggplot() affects all layers; a mapping inside a geom affects only that layer.
• A layer can override data (data =) to highlight or annotate a subset.
• inherit.aes = FALSE opts a layer out of the base mappings entirely.
• Add the layer you want on top last, since layers composite in order.