pairplot
Mojo module 🡭
pairplot
Every numeric variable against every other, in one grid (#353).
The first thing many people run on a new dataset, because it answers “what is related to what” in one figure.
This is composition, not a new mark. Cell (i, j) is a scatter of
variable j against variable i, the diagonal is variable i’s own
distribution, and the whole thing goes through render_facets().
What makes it a pairplot rather than a grid of unrelated charts is the
scales. Column j shares one x-domain down its whole height and row i
shares one y-domain across its whole width, both pinned from the
variable’s own range. Without that the cells are not comparable and the
figure lies: two panels of the same variable would draw it at two
different widths and a reader would read the difference as data.
Mark.CORRPLOT answers a related question and is not a substitute. A
correlation of 0.0 looks the same for independent data and for a perfect
parabola; the scatter is where you see which one you have.