Causal Inference in Observational Studies Using Counterfactual Estimation with Machine Learning Models
Keywords:
Causal inference, counterfactual estimation, observational studies, machine learning, treatment effect, potential outcomes, uplift modeling, propensity scoreAbstract
The increasing complexity of real-world data has amplified interest in applying causal inference techniques beyond randomized controlled trials. Observational studies, although rich in information, pose inherent challenges due to confounding variables and selection biases. Machine learning (ML) models, particularly counterfactual estimation methods, have emerged as powerful tools to infer causal effects by approximating the potential outcomes framework. This paper provides a comprehensive analysis of counterfactual estimation approaches for causal inference in observational studies using ML, considering contemporary advances and limitations. Literature, compare state-of-the-art methods, and illustrate their applications with visual models. Through theoretical discussions and empirical observations, we highlight the transformative potential of ML in making robust causal claims from observational data.
References
Hill, J. L. Bayesian nonparametric modeling for causal inference. J Comput Graph Stat, 20(1), 217–240.
Wager, S., & Athey, S. Estimation and inference of heterogeneous treatment effects using random forests. J Am Stat Assoc, 113(523), 1228–1242.
Louizos, C., et al. Causal effect inference with deep latent-variable models. NeurIPS, 3855–3865.
Shalit, U., Johansson, F. D., & Sontag, D. Estimating individual treatment effect: Generalization bounds and algorithms. ICML, 3076–3085.
Pearl, J. Causality: Models, Reasoning, and Inference. Cambridge University Press.
Rosenbaum, P. R., & Rubin, D. B. The central role of the propensity score in observational studies. Biometrika, 70(1), 41–55.
Johansson, F. D., Shalit, U., & Sontag, D. Learning representations for counterfactual inference. ICML, 3020–3029.
Athey, S., & Imbens, G. Recursive partitioning for heterogeneous causal effects. PNAS, 113(27), 7353–7360.
Van der Laan, M. J., & Rose, S. Targeted learning: Causal inference for observational and experimental data. Springer.
Funk, M. J., et al. Doubly robust estimation of causal effects. Am J Epidemiol, 173(7), 761–767.
King, G., & Nielsen, R. Why propensity scores should not be used for matching. Pol Anal, 27(4), 435–454.
Crump, R. K., et al. Dealing with limited overlap in estimation of average treatment effects. Biometrika, 96(1), 187–199.
Oberst, M., & Sontag, D. Counterfactual off-policy evaluation with Gumbel-max structural causal models. UAI, 2019.
Kennedy, E. H. Optimal doubly robust estimation of the average treatment effect. Res Synth Methods, 11(3), 420–435.
Lundberg, S. M., & Lee, S. I. A unified approach to interpreting model predictions. NeurIPS, 4765–4774.
