A perspective on covariate adjustment and confounding in randomized experiments.

by Darren1 min read25th Jan 2021No comments


World ModelingRationality

Randomized experiments (including clinical trials) are widely recognized as useful tools for trying to understand cause and effect. They are also criticized, both fairly and unfairly. Some of the more common misconceptions around randomization follow from the idea that randomization is useful because it "balances confounders", but this is not correct..not exactly anyway. The linked essay tries to explain why it's not correct, why it matters, and the implications this misconception has for the design of clinical trials. 


New Comment