Simpson's paradox
This is a very weird phenomenon that happens when the population count greatly affects the outcome of actual percentages. The following is some medical data from patients with mild and severe conditions treated in two different ways. You can see how the overall success rate turns out to be the other way around when the mild/severe groups are added together.
| Treatment A Successful | Treatment B Successful | |
|---|---|---|
| Mild Condition | 81/87 (93%) | 234/270 (87%) |
| Severe Condition | 192/263 (73%) | 55/80 (69%) |
| Overall | 273/350 (78%) | 289/350 (83%) |