A probabilistic or deterministic rule for selecting some units to receive control and other units to receive treatment.
It should be:
- ignorable: the assignment is independent of unobserved potential outcomes. In other words, we can ignore what would happen if the units receive the treatment or control. (by default, through randomization )
- unconfounded: the assignment is independent of ALL potential outcomes. It’s as if we’re making these decisions randomly, without being influenced by any variables, both the ones we know about and the ones we don’t.