Does a larger sample remove confounding?
No. A larger sample reduces sampling variation, but it does not repair systematic differences between treated and untreated people. If a common cause still influences both treatment and outcome, an unadjusted estimate can converge precisely to the wrong value.
Try it in the experiment
- Increase how strongly the risk score influences treatment.
- Compare the unadjusted estimate with the true effect.
- Open “Compare repeated studies.” The estimates become a stable pattern, but the pattern remains displaced from truth.
Valid design or adjustment can address measured common causes. Merely collecting more observations cannot recover information that the analysis ignores.