Generalizability and Transportability

In practical uses of causal inference methods, the goal is to not only estimate the causal effect of a treatment in the data in hand, but to use this estimate to infer the causal effect in a broader population. In medicine, for example, doctors or public health officials may want to know how to apply findings of a study to their population of patients or in their community. Or, in education or social welfare, school officials or policy-makers may want to know how to apply the findings to students in their schools, school district, or state. If unit specific treatment effects are all identical, these inferences from the sample to the population are straightforward. However, when treatment impacts vary, the causal effect estimated in a sample of data may not directly generalize to the target population of interest.