Ratio of variation between GROUPS to INDIVIDUAL VARIATION, if high, then group has effect. If people don't vary much at all but big group differences, prob not chance.
Goes back to central limit theorem.
Formulas--run them as code, logical. GOOD FOR WILL.