Optimal Design of Experiments in the Presence of Network Interference

• Activity Number: 670
• Type: Invited
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Edo Airoldi*
• Companies: Harvard University

Abstract

Causal inference research in statistics has been largely concerned with estimating the effect of treatment (e.g. personalized tutoring) on outcomes (e.g., test scores) under the assumption of "lack of interference"; that is, the assumption that the outcome of an individual does not depend on the treatment assigned to others. Moreover, whenever its relevance is acknowledged (e.g., study groups), interference is typically dealt with as an uninteresting source of variation in the data. In many modern applications, however, the lack of interference assumption is not tenable. Not only interference is present in these situations, and is an important aspect of the problem that cannot be abstracted away, but we are often interested in estimating the casual effect of such interference. In this talk, we will review challenges for causal analyses in the presence of interference. We will then introduce a two-stage strategy based on a piece-wise constant approximation of a graphon to define an optimal set of randomizations for estimating interference on large social and information networks.