Double machine learning for treatment and causal parameters

By Victor Chernozhukov, D. Chetverikov, Mert Demirer, Esther Duflo, Christian Hansen, Whitney K. Newey. 2016

200 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Statistics and Probability

In Causal Inference & Discovery

Citations in the corpus

Balanced Policy Evaluation and Learning (2017, methodology): “[13] propose a weighting-based approach that combines the direct and IPW approaches by adapting the doubly robust (DR) estimator [11, 34, 35, 38]: ⌧̂DR(⇡) = 1 n Pn i=1 Pm t=1 ⇡t(Xi)μ̂t(Xi) + 1 n Pn i=1(Yi μ̂Ti(Xi))⇡Ti(Xi)/'̂Ti(Xi).”
Generalized random forests (2016, methodology): “…the weights αi(x) are not sharply concentrated around x. Similar orthogonalization ideas have proven to be useful in many statistical contexts (e.g., Chernozhukov et al., 2016; Newey, 1994b; Neyman, 1979); in particular, Robinson (1988) showed that if we have access to a neighborhood S over which…”
Approximate residual balancing: debiased inference of average treatment effects in high dimensions (2016, methodology): “[2016] and Farrell [2015] in that these methods all require the treatment propensities e(x) estimable at an n−1/4 rate; and then consider various methods that can be used for estimation e(x), ranging from penalized regression [Farrell, 2015] to boosting [Chernozhukov et al., 2016].”
Efficient Policy Learning (2017, methodology): “There is a recent, active, literature that seeks to use tools from optimization and machine learning to develop semiparametrically efficient average treatment effect estimation with improved finite sample performance: examples include double machine learning (Belloni et al., 2017; Chernozhukov et al”
On the multiply robust estimation of the mean of the g-functional (2017, methodology): “1 we describe how to overcome this problem by splitting the sample and using a cross-fit estimator, a name coined in Chernozhukov, (2016). To obtain a cross-fit estimator we first randomly split the sample into U, say 5, equal sized subsamples u = 1, .”
Targeted Undersmoothing. (2017, methodology): “25The behavior may be improved by considering double machine learning as defined in [22], which relies on weaker sparsity conditions than [13].”
Estimating Average Treatment Effects: Supplementary Analyses and Remaining Challenges (2017, methodology): “Chernozhukov et al. [2016], in the context of much more general estimation problems, propose a closely related estimator focusing on the orthogonality properties arising from the use of the efficient score.”
Robust Estimation of Causal Effects via High-Dimensional Covariate Balancing Propensity Score. (2018, methodology): “2 considers a simple modification of the algorithm based on the sample splitting approach (Chernozhukov et al., 2016), which requires a weaker sparsity assumption (s1s2) 1/2 log(d∨ n)/n1/2 = o(1).”
Double/de-biased machine learning using regularized Riesz representers (2018, methodology): “…γ 0 and α 0 around their true values (see Lemma 2 below for details), a property sometime referred to as the Neyman-type orthogonality (see, e.g., Chernozhukov et al. (2016a)), making this representation a good one to use as a basis for estimation and inference in modern high-dimensional…”
Estimating individual treatment effect: generalization bounds and algorithms (2016, background): “However, most previous work on this subject focused on asymptotic consistency (Belloni et al., 2014; Athey et al., 2016; Chernozhukov et al., 2016), and so far there has not been much work on the generalization-error of such a procedure.”