See the effect size: overlap, superiority, group size needed
In brief
The effect size d says by how many standard deviations two groups differ on average. With the slider one can see how two bell curves move apart: at d = 0.2 they lie almost on top of each other, at d = 0.8 clearly shifted but still overlapping.
What the numbers mean
- Overlap: the share of the area that both curves have in common.
- Share above the mean of the other group (Cohen's U3): how many of the "better" group lie above the average of the other?
- Probability of superiority: if one person is drawn from each group, how often does the one from the "better" group have the higher value? At d = 0 it is 50%.
- Required group size: approximation for a two-sided test with α = 5% and 80% power; the page "Plan a study" calculates more exactly.
Caution
"Small", "medium" and "large" are conventions (Cohen 1988) for the behavioural sciences. In medicine d = 0.2 can be very important, in education d = 0.4 can already be a lot. The pictures hold for two equally spread normal distributions; with skewed data they look different.