Learning Overlapping Representations for the Estimation of Individualized Treatment Effects

Abstract

The choice of making an intervention depends on its potential benefit or harm in comparison to alternatives. Estimating the likely outcome of alternatives from observational data is a challenging problem as all outcomes are never observed, and selection bias precludes the direct comparison of differently intervened groups. Despite their empirical success, we show that algorithms that learn domain-invariant representations of inputs (on which to make predictions) are often inappropriate, and develop generalization bounds that demonstrate the dependence on domain overlap and highlight the need for invertible latent maps. Based on these results, we develop a deep kernel regression algorithm and posterior regularization framework that substantially outperforms the state-of-the-art on a variety of benchmarks data sets.

References in corpus (3)

• Density estimation using Real NVP
• On Learning Invariant Representation for Domain Adaptation
• Learning Weighted Representations for Generalization Across Designs

Cited by in corpus (4)

• Robust Recursive Partitioning for Heterogeneous Treatment Effects with Uncertainty Quantification
• Regret Minimization for Causal Inference on Large Treatment Space
• Doing Great at Estimating CATE? On the Neglected Assumptions in Benchmark Comparisons of Treatment Effect Estimators
• Neuroevolutionary Feature Representations for Causal Inference