Cartesian and Beyond

The coordinate system is the space your marks are placed in. You have been using one all along — the standard Cartesian plane, x running left-to-right and y bottom-to-top — without naming it. Making it an explicit component unlocks zooming, aspect-ratio control, and (next page) entirely different chart shapes.

The default: Cartesian

coord_cartesian() is the familiar flat plane. It is applied automatically, so these two are identical:

Coordinates transform the space, not the data

Here is the key mental model. Scales and stats operate in data space — they change values. Coordinates operate on the final space — they reshape where already-computed marks land, after the stats and scales have run.

This ordering is exactly why coord_cartesian(xlim = ...) zooms rather than filters: the coordinate step happens last, after the smoother has already been fit on all the data. (We met this on the Position Scales page — now you can see why.)

Fixing the aspect ratio

coord_fixed() forces one data unit on x to equal one data unit on y — essential when x and y are in the same units and distances should be comparable (maps, before/after on the same measurement scale):

With coord_fixed(), the red y = x line is a genuine 45° diagonal, so you can read "highway beats city MPG" as every point above the line. Without it, the plot would stretch to fill the panel and distort that comparison.

QuestionSelect one

In the ggplot pipeline, when does the coordinate system act relative to stats and scales?

Before the stat, so it can change which rows are counted.

After stats and scales, arranging the already-computed marks in the final space — which is why coord_cartesian(xlim=) zooms instead of filtering.

At the same time as the mapping.

Coordinates never affect a plot; they are decorative.

QuestionSelect one

Why might you add coord_fixed() to a plot of city MPG vs. highway MPG?

To filter out cars with low MPG.

To convert the axes to a log scale.

Because both axes are in the same units, so equal data distances should look equal — making a y = x reference line a true 45-degree diagonal.

To add a confidence band to the points.

Key takeaways

• The coordinate system is the space marks occupy; Cartesian is the default (coord_cartesian()).
• Coordinates act last, on the final space — so coord_cartesian(xlim=) zooms without discarding data.
• coord_fixed() enforces an equal x/y unit ratio, vital when the axes share units.
• Coordinates are a genuine, swappable component — the foundation for the polar and flipped charts on the next page.