Negative control outcome
= Negative control outcome
A negative control outcome cannot plausibly be caused by treatment, so an estimated effect can reveal bias.
Unlike treatment comparison alone, this <scientific control> checks for an association that should vanish under the intended causal model.