Rubin Causal Model
摘要
The first part is conceptual and implies that we should always start an inquiry into a causal question by carefully defining all causal estimands (quantities to be estimated) in terms of potential outcomes, which are all values that could be observed in some real or hypothetical experiment comparing the results under a well-defined active treatment versus the results under a well-defined control treatment in a population of units, each of which can be exposed, in principle, to either treatment.