Causal Effect Inference for Structured Treatments
This addresses causal inference for complex data types, which is important for fields like medicine and social sciences, though it appears incremental as it builds on existing CATE estimation frameworks.
The paper tackles the problem of estimating conditional average treatment effects (CATEs) for structured treatments like graphs, images, and texts, proposing a generalized Robinson decomposition that reduces regularization bias and achieves quasi-oracle convergence, with experiments on small-world and molecular graphs showing it outperforms prior methods.
We address the estimation of conditional average treatment effects (CATEs) for structured treatments (e.g., graphs, images, texts). Given a weak condition on the effect, we propose the generalized Robinson decomposition, which (i) isolates the causal estimand (reducing regularization bias), (ii) allows one to plug in arbitrary models for learning, and (iii) possesses a quasi-oracle convergence guarantee under mild assumptions. In experiments with small-world and molecular graphs we demonstrate that our approach outperforms prior work in CATE estimation.