Using Embeddings to Correct for Unobserved Confounding in Networks
Abstract
We consider causal inference in the presence of unobserved confounding. We study the case where a proxy is available for the unobserved confounding in the form of a network connecting the units. For example, the link structure of a social network carries information about its members. We show how to effectively use the proxy to do causal inference. The main idea is to reduce the causal estimation problem to a semi-supervised prediction of both the treatments and outcomes. Networks admit high-quality embedding models that can be used for this semi-supervised prediction. We show that the method yields valid inferences under suitable (weak) conditions on the quality of the predictive model. We validate the method with experiments on a semi-synthetic social network dataset. Code is available at github.com/vveitch/causal-network-embeddings.
Cite
@article{arxiv.1902.04114,
title = {Using Embeddings to Correct for Unobserved Confounding in Networks},
author = {Victor Veitch and Yixin Wang and David M. Blei},
journal= {arXiv preprint arXiv:1902.04114},
year = {2019}
}
Comments
An earlier version also addressed the use of text embeddings. That material has been expanded and moved to arxiv:1905.12741, "Using Text Embeddings for Causal Inference"