A principal stratum is defined by the joint potential outcomes of an intermediate variable under all compared treatments, rather than by its one observed value. For binary assignment and intermediate response , a basic principal stratum is the set of units with a specified pair . Each unit belongs to the same principal stratum whichever assignment it receives, although only is observed.[ref]
Principal Stratification in Causal Inference by Constantine E. Frangakis and Donald B. Rubin. Summary and section 2: cross-classification by joint intermediate potential values
For example, let record whether a unit takes a treatment when offered encouragement. The pairs , , and define never-takers, compliers, defiers and always-takers, respectively. An instrumental-variable monotonicity assumption rules out defiers. A comparison conditional on observed receipt generally mixes compliers with always-takers and can mix them in different proportions across assignment arms. It therefore does not automatically identify a causal effect within either principal stratum.
New to topics? Read the docs here!