Assessing Principal Causal Effects Using Principal Score Methods

We first introduce the notation and the basic framework of principal stratification. Then, we review some identifying (structural) assumptions – ignorability of the treatment assignment mechanism, monotonicity, exclusion restrictions and principal ignorability assumptions – and discuss their plausibility in different settings. We define principal scores, discuss their properties and provide some details on the estimator under monotonicity. We review identification and estimation strategies of principal causal effects based on weighting on principal score under monotonicity and principal ignorability assumptions. We deal with the specification of the principal score model. We review and discuss methods to conducting sensitivity analysis with respect to principal ignorability and monotonicity. We offer some discussion on possible extensions of principal stratification analysis based on principal scores under principal ignorability together with some concluding remarks.