Confounding from a common cause

Change how strongly the risk score influences treatment and compare the estimated effect with the known true effect.

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

  1. Increase how strongly the risk score influences treatment.
  2. Compare the unadjusted estimate with the true effect.
  3. 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.