Propensity Score Matching for Causal Inference with Relational Data

This work provides an automated method to derive non-intuitive relational variables based on the relational structure and a set of naive confounders and shows that this method improves the performance of PSM for causal inference with relational data.