Abstract

Defining estimands for indirect, total, and overall effects of an intervention in the potential outcomes framework of causal inference is not straightforward.  Often in causal inference the assumption of no interference is made wherein the outcome in one individual is independent of the treatment assignment in the other individuals in the population (Cox 1958).  However, in many contexts, for example infectious diseases and vaccination, this assumption does not hold. Indeed, interference is the source of the indirect, total, overall and other effects of interest. In recent years, methods and applications for causal inference with interference have expanded rapidly. We discuss some approaches to estimating effects of interest under interference, including the assumption of partial interference, allowing each individual to have its own interference set, and networks. We present an example of estimating direct, indirect, total, and overall effects of vaccination. We discuss exciting directions of the field of causal inference with interference.