Estimating individual treatment effect: generalization bounds and algorithms
arXiv:1606.03976
Abstract
There is intense interest in applying machine learning to problems of causal inference in fields such as healthcare, economics and education. In particular, individual-level causal inference has important applications such as precision medicine. We give a new theoretical analysis and family of algorithms for predicting individual treatment effect (ITE) from observational data, under the assumption known as strong ignorability. The algorithms learn a "balanced" representation such that the induced treated and control distributions look similar. We give a novel, simple and intuitive generalization-error bound showing that the expected ITE estimation error of a representation is bounded by a sum of the standard generalization-error of that representation and the distance between the treated and control distributions induced by the representation. We use Integral Probability Metrics to measure distances between distributions, deriving explicit bounds for the Wasserstein and Maximum Mean Discrepancy (MMD) distances. Experiments on real and simulated data show the new algorithms match or outperform the state-of-the-art.
arXiv commentAdded name "TARNet" to refer to version with alpha = 0. Removed supp
References in corpus (12)
- Adam: A Method for Stochastic Optimization
- Recursive Partitioning for Heterogeneous Causal Effects
- Frustratingly Easy Domain Adaptation
- Domain Adaptation: Learning Bounds and Algorithms
- Fast Computation of Wasserstein Barycenters
- Learning with a Wasserstein Loss
- Distinguishing cause from effect using observational data: methods and benchmarks
- Stochastic Optimization for Large-scale Optimal Transport
- Learning from Logged Implicit Exploration Data
- Constraint-based Causal Discovery from Multiple Interventions over Overlapping Variable Sets
- Double/Debiased Machine Learning for Treatment and Causal Parameters
- Counterfactual Prediction with Deep Instrumental Variables Networks
Cited by in corpus (17)
- Discussion on "Bayesian Regression Tree Models for Causal Inference: Regularization, Confounding, and Heterogeneous Effects" by Hahn, Murray and Carvalho
- Bayesian Inference of Individualized Treatment Effects using Multi-task Gaussian Processes
- Learning Counterfactual Representations for Estimating Individual Dose-Response Curves
- Adapting Neural Networks for the Estimation of Treatment Effects
- Support and Invertibility in Domain-Invariant Representations
- An Evaluation Toolkit to Guide Model Selection and Cohort Definition in Causal Inference
- Invariant Representation Learning for Treatment Effect Estimation
- Estimation of Individual Treatment Effect in Latent Confounder Models via Adversarial Learning
- Reinforcement Learning for Uplift Modeling
- Inferring Individual Direct Causal Effects Under Heterogeneous Peer Influence
- Nonparametric Heterogeneous Treatment Effect Estimation in Repeated Cross Sectional Designs
- Learning causal effects from many randomized experiments using regularized instrumental variables
- Rarely-switching linear bandits: optimization of causal effects for the real world
- Weighted Tensor Completion for Time-Series Causal Inference
- Neuroevolutionary Feature Representations for Causal Inference
- Hölder Bounds for Sensitivity Analysis in Causal Reasoning
- Learning and Testing Sub-groups with Heterogeneous Treatment Effects:A Sequence of Two Studies