B1046
Title: Nearly exact matching in the presence of networks
Authors: Alexander Volfovsky - Duke University (United States) [presenting]
Abstract: A classical problem in causal inference is that of matching treatment units to control units in an observational dataset. This problem is distinct from simple estimation of treatment effects as it provides additional practical interpretability of the underlying causal mechanisms that is not available without matching. Some of the main challenges in developing matching methods arise from the tension among the desire for granular and interpretable matched groups while having enough data to learn causal effects while dealing with complicating factors such as networks and non-independence among units. To deal with the influence of networks we propose to learn which network components are relevant to our causal questions. We propose several optimization objectives for match quality that capture covariates and structures that are integral for making causal statements while encouraging as many matches as possible.
