Rank correlation (rs, τb)

A rank correlation compares not the numbers themselves but their rank positions. It suits columns with only a few levels or with an order only. Spearman (rs) and Kendall (τb) lie between −1 and +1 like r; for an equally strong relationship Kendall's τb comes out smaller.

Example: Coffee and concentration: rs = −.12 (120 people). Those who drink more coffee tend to score somewhat lower. The relationship is weak.

Caution: A relationship is not a cause. A rank correlation picks up any monotonic (consistently rising or falling) relationship, not only straight-line ones - a scatter plot shows the shape.