When you relate a multi-select to a categorical variable, Statwing runs a bunch of analyses under the hood and summarizes them into a table. Specifically, Statwing implicitly builds many tables, one for each checkbox option, and then takes the statistical results of the adjusted residuals (the little arrows) and inserts them into the table.

By way of example, let's say you have a dataset that looks like this:


If you related just one of those pet columns to the color column, you'd get a table like this:

Note that those arrows indicate that owning a cat is positively correlated with liking red and negatively correlated with liking Blue.

If you relate the overall pet checkbox group with the color variable, you'd get a table like this:


Note that the arrows show the same relationships between cat and both red and blue. That's because they're pulled from the very same analysis.

A helpful way to look at this analysis is to sort by the second column, in this case pet:

This view helps you clearly see that red is highly associated with cat, but not green.