Estimating individual treatment effect: generalization bounds and algorithms

2016/06/13 by Uri Shalit, Shalit, Uri, Fredrik Johansson +4 · 21 citations
Computer Science · Economics, Econometrics and Finance · Mathematics · #Advanced Causal Inference Techniques #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Health Systems, Economic Evaluations, Quality of Life #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Statistical Methods and Bayesian Inference #Statistical Methods and Inference #cs.AI #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.1606.03976

Added name "TARNet" to refer to version with alpha = 0. Removed supp

openalex publication_date 2016/06/13 · arxiv created 2017/05/16 · arxiv updated 2017/05/17 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

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.

Cited by

• DoWhy: Addressing Challenges in Expressing and Validating Causal Assumptions
• Adaptive Multi-Source Causal Inference
• Learning Continuous Treatment Policy and Bipartite Embeddings for Matching with Heterogeneous Causal Effects
• Active Learning for Decision-Making from Imbalanced Observational Data
• Assessing Disparate Impacts of Personalized Interventions: Identifiability and Bounds
• MultiMBNN: Matched and Balanced Causal Inference with Neural Networks
• Balance Regularized Neural Network Models for Causal Effect Estimation
• Identifying Causal-Effect Inference Failure with Uncertainty-Aware Models
• Learning Optimal Policies from Observational Data
• On the Distinction Between "Conditional Average Treatment Effects" (CATE) and "Individual Treatment Effects" (ITE) Under Ignorability Assumptions
• Predicting Individual Treatment Effects of Large-scale Team Competitions in a Ride-sharing Economy
• Towards unique and unbiased causal effect estimation from data with hidden variables
• RealCause: Realistic Causal Inference Benchmarking
• MetaCI: Meta-Learning for Causal Inference in a Heterogeneous Population
• Treatment Targeting by AUUC Maximization with Generalization Guarantees
• Split: Inferring Unobserved Event Probabilities for Disentangling Brand-Customer Interactions
• Estimating Treatment Effects with Observed Confounders and Mediators
• General Control Functions for Causal Effect Estimation from Instrumental Variables
• Improving Fair Predictions Using Variational Inference In Causal Models
• Sequential Deconfounding for Causal Inference with Unobserved Confounders
• Learning Weighted Representations for Generalization Across Designs

Related

• Generalization Bounds and Representation Learning for Estimation of\n Potential Outcomes and Causal Effects
• Estimating Individual Treatment Effects through Causal Populations\n Identification
• Learning Representations of Instruments for Partial Identification of Treatment Effects
• Individual Treatment Effect: Prediction Intervals and Sharp Bounds
• Causes of Effects: Learning individual responses from population data