Estimation and Inference of Heterogeneous Treatment Effects using Random Forests
By Stefan Wager, Susan Athey. 2015
3,108 citations recorded by the source corpus.
ABSTRACT Many scientific and engineering challenges—ranging from personalized medicine to customized marketing recommendations—require an understanding of treatment effect heterogeneity. In this article, we develop a nonparametric causal forest for estimating heterogeneous treatment effects that extends Breiman’s widely used random forest algorithm. In the potential outcomes framework with unconfoundedness, we show that causal forests are pointwise consistent for the true treatment effect and have an asymptotically Gaussian and centered sampling distribution. We also discuss a practical method for constructing asymptotic confidence intervals for the true treatment effect that are centered at the causal forest estimates. Our theoretical results rely on a generic Gaussian theory for a large family of random forest algorithms. To our knowledge, this is the first set of results that allows any type of random forest, including classification and regression forests, to be used for provably valid statistical inference. In experiments, we find causal forests to be substantially more powerful than classical methods based on nearest-neighbor matching, especially in the presence of irrelevant covariates.
Subfields: Statistics and Probability, Causal Inference & Discovery
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Statistics and Probability
- Stefan Wager
- Susan Athey
- Victor Chernozhukov
- Guido W. Imbens
- Cynthia Rudin
- Alexandre Belloni
- Christian Hansen
- Mihaela van der Schaar
In Causal Inference & Discovery
- Susan Athey
- Mihaela van der Schaar
- Stefan Wager
- Ahmed M. Alaa
- Guido W. Imbens
- Ioana Bica
- Cynthia Rudin
- Victor Chernozhukov
Citations in the corpus
- Machine Learning Methods That Economists Should Know About (2019, background): “These will rapidly degrade the performance of kernel regression, but will not affect random forest nearly as severely because it will largely ignore them [Wager and Athey, 2017].”
- Time Series Deconfounder: Estimating Treatment Effects over Time in the Presence of Hidden Confounders (2019, methodology): “Most of the previous methods for performing causal inference have focused on the static setting (Hill, 2011; Wager & Athey, 2017; Alaa & van der Schaar, 2017; Yoon et al., 2018), and less attention has been given to the time series setting.”
- Deep-Treat: Learning Optimal Personalized Treatments From Observational Data Using Neural Networks (2018, methodology): “• Causal Forest is a random forest variant for causal inference (Wager and Athey 2015).”
- FLAME: A Fast Large-scale Almost Matching Exactly Approach to Causal Inference (2017, methodology): “We compared FLAME with several other methods including: (1) one-to-one Propensity Score Nearest Neighbor Matching (1-PSNNM) (Ross et al., 2015), (2) Genetic Matching (GenMatch) (Diamond and Sekhon, 2013), (3) Causal Forest (Wager and Athey, 2017), (4) Mahalanobis Matching and (5) double linear regre”
- Deep Neural Networks for Estimation and Inference: Application to Causal Effects and Other Semiparametric Estimands (2018, methodology): “, 2011, 2012; Farrell, 2015) and tree/forest based methods (Wager and Athey, 2018), though earlier work studied shallow (typically with a single hidden layer) neural networks with smooth activation functions (White, 1989, 1992; Chen and White, 1999).”
- Local Linear Forests (2018, methodology): “…linear forests for nonparametric regression; however, a similar local linear correction can also be applied to quantile regression forests (Meinshausen 2006), causal forests (Wager and Athey 2018) or, more broadly, to any instance of generalized random forests (Athey, Tibshirani, and Wager 2019).”
- Metalearners for estimating heterogeneous treatment effects using machine learning (2017, methodology): “This version of RF is particularly accessible from a theoretical point of view; it performs well in noisy settings; and it is better suited for inference (8, 17).”
- Counterfactual Prediction with Deep Instrumental Variables Networks (2016, methodology): “Splitting procedures like this have a long history [Cox, 1975], and have recently regained popularity as an option for inference post model-selection [e.g., Fithian et al., 2015, Wager and Athey, 2015].”
- Generalized random forests (2016, methodology): “(47) As in arguments used by Wager and Athey (2018), we see that our quantity of interest U -statistic over T , and in particular δ ( (θ, ν) , ( θ′, ν ′ )) = ( n bs/2c, ds/2e )−1 ∑ {S1,S2∈{1, ..., n}} Tθ, ν(J1, J2)− Tθ′, ν′(J1, J2).”
- The Econometrics of Randomized Experiments (2016, methodology): “Recently, Wager and Athey (2015) propose a modified version of the random forest algorithm that produces treatment effect estimates that can be shown to be asymptotically normal and centered on the true value of the treatment effect, and they propose a consistent estimator for the asymptotic…”
- Recursive partitioning for heterogeneous causal effects (2015, background): “In some contexts, additional assumptions such as “sparsity” (only a few covariates affect the outcomes) can be applied to guarantee consistency or asymptotic normality of predictions ([28]).”
- Deep Counterfactual Networks with Propensity-Dropout (2017, methodology): “More recent works tackled the problem of estimating “indi-vidualized” treatment effects using representation learning (Johansson et al., 2016; Shalit et al., 2017), Bayesian inference (Hill, 2012), and standard supervised learning (Wager & Athey, 2015).”
- Learning When-to-Treat Policies (2019, background): “The problem of learning optimal decision rules is also closely related to learning heterogeneous treatment effects (Athey and Imbens, 2016; Athey, Tibshirani, and Wager, 2019; Chen, 2007; Künzel, Sekhon, Bickel, and Yu, 2019; Nie and Wager, 2017; Wager and Athey, 2018).”
- Learning Representations for Counterfactual Inference (2016, methodology): “These methods include any type of regression including linear and logistic regression (Pren-tice, 1976; Gelman & Hill, 2006), random forests (Wager & Athey, 2015) and regression trees (Chipman et al., 2010).”
- Representation Balancing MDPs for Off-Policy Policy Evaluation (2018, neutral): “Recently, different kinds of machine learning 97 models such as Gaussian Processes [1], random forests [24], and GANs [8] have been used to estimate 98 heterogeneous treatment effects (HTE), in non-sequential settings.”
- Efficient Policy Learning (2017, background): “Finally, we also note a growing literature on estimating conditional average treatment effects (Athey and Imbens, 2016; Athey, Tibshirani, and Wager, 2019; Chen, 2007; Künzel, Sekhon, Bickel, and Yu, 2017; Nie and Wager, 2017; Wager and Athey, 2018).”