Simpson’s Paradox Mixing Lab
This interactive lab uses two surgeons as the running example. The within-group success rates are fixed in the block below. Use the sliders to choose the case mix for Surgeon A and Surgeon B, then watch whether a Simpson reversal appears after the patient groups are mixed.
Within each risk group
Surgeon A is better in both groups. This part never changes.
After mixing groups
The overall rates are weighted averages. Changing the group mix changes the weights, not the within-group success rates.