A0999
Title: Unconfoundedness with network interference
Authors: Michael Leung - UC Santa Cruz (United States) [presenting]
Abstract: The aim is to study the nonparametric estimation of treatment and spillover effects using observational data from a single large network. A model of network interference is considered that allows for peer influence in outcomes and selection into treatment but requires influence to decay with network distance. In this setting, the network and covariates of all units can be potential sources of confounding, in contrast to existing work that assumes confounding is limited to a known, low-dimensional function of these objects. To estimate the first-stage nuisance functions of the doubly robust estimator, it is proposed to use neural graph networks, which are designed to approximate functions of graph-structured inputs. Under the proposed model of interference, primitive conditions for a network analogue of approximate sparsity are derived, which provides justification for the use of shallow architectures.
