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.