mlsanityData to 14 Oct 2026

RCT Rejection Sampling for Causal Estimation Evaluation

arXiv 2307.15176 ↗First posted 27 Jul 2023cs.AI cs.CL cs.LGcited by 7

Not on the map. Why?

Abstract (arXiv)

Confounding is a significant obstacle to unbiased estimation of causal effects from observational data. For settings with high-dimensional covariates -- such as text data, genomics, or the behavioral social sciences -- researchers have proposed methods to adjust for confounding by adapting machine learning methods to the goal of causal estimation. However, empirical evaluation of these adjustment methods has been challenging and limited. In this work, we build on a promising empirical evaluation strategy that simplifies evaluation design and uses real data: subsampling randomized controlled trials (RCTs) to create confounded observational datasets while using the average causal effects from the RCTs as ground-truth. We contribute a new sampling algorithm, which we call RCT rejection sampling, and provide theoretical guarantees that causal identification holds in the observational data to allow for valid comparisons to the ground-truth RCT. Using synthetic data, we show our algorithm indeed results in low bias when oracle estimators are evaluated on the confounded samples, which is not always the case for a previously proposed algorithm. In addition to this identification result, we highlight several finite data considerations for evaluation designers who plan to use RCT rejection sampling on their own datasets. As a proof of concept, we implement an example evaluation pipeline and walk through these finite data considerations with a novel, real-world RCT -- which we release publicly -- consisting of approximately 70k observations and text data as high-dimensional covariates. Together, these contributions build towards a broader agenda of improved empirical evaluation for causal estimation.

Key facts. Cited by 7 papers in the map (data through 14 Oct 2026). Builds on 26 of its 67 references that are in the map. First posted on arXiv 27 Jul 2023.

Builds on

26 of 67 references.

References · 26

  1. 2019Do ImageNet Classifiers Generalize to ImageNet?cited by 1,515
  2. 2017Meta-learners for Estimating Heterogeneous Treatment Effects using Machine Learningcited by 478
  3. 2016Learning Representations for Counterfactual Inferencecited by 329
  4. 2019Adapting Neural Networks for the Estimation of Treatment Effectscited by 287
  5. 2018CatBoost: gradient boosting with categorical features supportcited by 208
  6. 2021Causal Inference in Natural Language Processing: Estimation, Prediction, Interpretation and Beyondcited by 178
  7. 2021Beware of the Simulated DAG! Causal Discovery Benchmarks May Be Easy To Gamecited by 161
  8. 2017Automated versus do-it-yourself methods for causal inference: Lessons learned from a data analysis competitioncited by 106
  9. 2018Removing Hidden Confounding by Experimental Groundingcited by 64
  10. 2018Deep Neural Networks for Estimation and Inferencecited by 62
  11. 2020Text and Causal Inference: A Review of Using Text to Remove Confounding from Causal Estimatescited by 59
  12. 2016Approximate Residual Balancing: De-Biased Inference of Average Treatment Effects in High Dimensionscited by 50
14 more
  1. 2020RealCause: Realistic Causal Inference Benchmarkingcited by 46
  2. 2018Benchmarking Framework for Performance-Evaluation of Causal Inference Analysiscited by 33
  3. 2018Challenges of Using Text Classifiers for Causal Inferencecited by 33
  4. 2019The Case for Evaluating Causal Models Using Interventional Measures and Empirical Datacited by 30
  5. 2020Semiparametric Inference For Causal Effects In Graphical Models With Hidden Variablescited by 23
  6. 2022Validating Causal Inference Methodscited by 20
  7. 2020Adjusting for Confounders with Text: Challenges and an Empirical Evaluation Framework for Causal Inferencecited by 13
  8. 2020How and Why to Use Experimental Data to Evaluate Methods for Observational Causal Inferencecited by 13
  9. 2021Generating Synthetic Text Data to Evaluate Causal Inference Methodscited by 9
  10. 2021Deconfounding Scores: Feature Representations for Causal Effect Estimation with Weak Overlapcited by 9
  11. 2022On Testability of the Front-Door Model via Verma Constraintscited by 8
  12. 2017Bias and high-dimensional adjustment in observational studies of peer effectscited by 7
  13. 2021Statistical Testing under Distributional Shiftscited by 6
  14. 2022Randomized trials and their observational emulations: a framework for benchmarking and joint analysiscited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 1; 2024: 3; 2025: 1; 2026: 2.

7 papers.

  1. 2024End-To-End Causal Effect Estimation from Unstructured Natural Language Datacited by 18
  2. 2023Hidden yet quantifiable: A lower bound for confounding strength using randomized trialscited by 8
  3. 2024Proximal Causal Inference With Text Datacited by 7
  4. 2025Position: Causal Machine Learning Requires Rigorous Synthetic Experiments for Broader Adoptioncited by 6
  5. 2026Using Text-Based Causal Inference to Disentangle Factors Influencing Online Review Ratingscited by 1
  6. 2024Network Causal Effect Estimation In Graphical Models Of Contagion And Latent Confoundingcited by 0
  7. 2026Exploring Sparse Autoencoders in Text-Based Causal Confounding Adjustmentcited by 0