Approximate Residual Balancing: Debiased Inference of Average Treatment Effects in High Dimensions

Athey, Susan; Imbens, Guido W.; Wager, Stefan

doi:10.1111/rssb.12268

Cited by 281 publications

(300 citation statements)

References 85 publications

Supporting

Mentioning

295

Contrasting

Order By: Relevance

“…Within this setting, there is a large classical literature focused on low-dimensional settings that provides methods for adjusting for confounding variables including regression methods, propensity score adjustment methods, matching methods, and "doubly-robust" combinations of these methods; see, for example, Robins and Rotnitzky (1995), Hahn (1998), Hirano et al (2003), and Abadie and Imbens (2006) as well as the textbook overview provided in Imbens and Rubin (2015). In this section, we present results that complement this important classic work as well as the rapidly expanding body of work on estimation under unconfoundedness using ML methods; see, among others, Athey et al (2016), , , Farrell (2015), and Imai and Ratkovic (2013). We specifically consider estimation of average treatment effects when treatment effects are fully heterogeneous and the treatment variable is binary, D ∈ {0, 1}.…”

Section: Inference On Ate and Attesupporting

confidence: 52%

“…We note that similar refined rates have appeared in the context of estimation of treatment effects in high-dimensional settings under sparsity; see Farrell (2015) and Athey et al (2016) and related discussion in Remark 5.2. Our refined rate results complement this work by applying to a broad class of estimation contexts, including estimation of average treatment effects, and to a broad set of ML estimators.…”

Section: This Occurs In the Following Important Examplesmentioning

confidence: 98%

“…We note that similar refined rates appeared in Farrell (2015) who considers estimation of treatment effects in a setting where an approximately sparse model holds for both the regression and propensity score functions. In interesting related work, Athey et al (2016) show that √ N consistent estimation of an average treatment effect is possible under very weak conditions on the propensity score -allowing for the possibility that the propensity score may not be consistently estimated -under strong sparsity of the regression function such that s g √ N . Thus, the approach taken in this context and Athey et al (2016) are complementary and one may prefer either depending on whether or not the regression function can be estimated extremely well based on a sparse method.…”

Section: Inference On Ate and Attementioning

confidence: 99%

See 2 more Smart Citations

Double/debiased machine learning for treatment and structural parameters

et al. 2017

View full text Add to dashboard Cite

Summary We revisit the classic semiparametric problem of inference on a low dimensional parameter θ0 in the presence of high-dimensional nuisance parameters η0. We depart from the classical setting by allowing for η0 to be so high-dimensional that the traditional assumptions, such as Donsker properties, that limit complexity of the parameter space for this object break down. To estimate η0, we consider the use of statistical or machine learning (ML) methods which are particularly well-suited to estimation in modern, very high-dimensional cases. ML methods perform well by employing regularization to reduce variance and trading off regularization bias with overfitting in practice. However, both regularization bias and overfitting in estimating η0 cause a heavy bias in estimators of θ0 that are obtained by naively plugging ML estimators of η0 into estimating equations for θ0. This bias results in the naive estimator failing to be N −1/2 consistent, where N is the sample size. We show that the impact of regularization bias and overfitting on estimation of the parameter of interest θ0 can be removed by using two simple, yet critical, ingredients: (1) using Neyman-orthogonal moments/scores that have reduced sensitivity with respect to nuisance parameters to estimate θ0, and (2) making use of cross-fitting which provides an efficient form of data-splitting. We call the resulting set of methods double or debiased ML (DML). We verify that DML delivers point estimators that concentrate in a N −1/2 -neighborhood of the true parameter values and are approximately unbiased and normally distributed, which allows construction of valid confidence statements. The generic statistical theory of DML is elementary and simultaneously relies on only weak theoretical requirements which will admit the use of a broad array of modern ML methods for estimating the nuisance parameters such as random forests, lasso, ridge, deep neural nets, boosted trees, and various hybrids and ensembles of these methods. We illustrate the general theory by applying it to provide theoretical properties of DML applied to learn the main regression parameter in a partially linear regression model, DML applied to learn the coefficient on an endogenous variable in a partially linear instrumental variables model, DML applied to learn the average treatment effect and the average treatment effect on the treated under unconfoundedness, and DML applied to learn the local average 1 Note: Replication files in programming language R are available at https://github.com/VC2015/DMLonGitHub/ 2 CCDDHNR treatment effect in an instrumental variables setting. In addition to these theoretical applications, we also illustrate the use of DML in three empirical examples.

show abstract

Section: Inference On Ate and Attesupporting

confidence: 52%

Section: This Occurs In the Following Important Examplesmentioning

confidence: 98%

Section: Inference On Ate and Attementioning

confidence: 99%

See 1 more Smart Citation

Double/debiased machine learning for treatment and structural parameters

et al. 2017

View full text Add to dashboard Cite

show abstract

“…Note that the procedure presented here can also be considered an alternative to causal estimation techniques based on propensity scores (e.g., Rosenbaum and Rubin, 1983; Little and Rubin, 2000). While adapting propensity score techniques to high-dimensional settings is an active area of research (e.g., Athey et al, 2016), it is typical to assume unconfoundedness (a.k.a. “strong ignorability”).…”

Section: Introductionmentioning

confidence: 99%

Estimating bounds on causal effects in high-dimensional and possibly confounded systems

Malinsky

Spirtes

2017

International Journal of Approximate Reasoning

View full text Add to dashboard Cite

We present an algorithm for estimating bounds on causal effects from observational data which combines graphical model search with simple linear regression. We assume that the underlying system can be represented by a linear structural equation model with no feedback, and we allow for the possibility of latent confounders. Under assumptions standard in the causal search literature, we use conditional independence constraints to search for an equivalence class of ancestral graphs. Then, for each model in the equivalence class, we perform the appropriate regression (using causal structure information to determine which covariates to adjust for) to estimate a set of possible causal effects. Our approach is based on the IDA procedure of Maathuis et al. (2009), which assumes that all relevant variables have been measured (i.e., no latent confounders). We generalize their work by relaxing this assumption, which is often violated in applied contexts. We validate the performance of our algorithm in simulation experiments.

show abstract

“…41 Recent investigations of penalized regression propensity matching also show a reduction in bias. 42,43 We believe our implementation reduced bias, as our estimate of the effect of CDI on LOS demonstrated significant deviations from unmatched analyses and concordance with the multistate matching analysis (which did not leverage propensity scores or matching). We also note again that propensity matched estimates offer a conservative effect size, which was the intention of this study.…”

Section: Discussionmentioning

confidence: 65%

Estimating Local Costs Associated WithClostridium difficileInfection Using Machine Learning and Electronic Medical Records

Pak

Chacko

O’Donnell

et al. 2017

Infect. Control Hosp. Epidemiol.

View full text Add to dashboard Cite

Background Reported per-patient costs of Clostridium difficile infection (CDI) vary by two orders of magnitude among different hospitals, implying that infection control officers need precise, local analyses to guide rational decision-making between interventions. Objective We sought to comprehensively estimate changes in length of stay (LOS) attributable to CDI at one urban tertiary-care facility using only data automatically extractable from the electronic medical record (EMR). Methods We performed a retrospective cohort study of 171,938 visits spanning a 7-year period. 23,968 variables were extracted from EMR data recorded within 24 hours of admission to train elastic net regularized logistic regression models for propensity score matching. To address time-dependent bias (reverse causation), we separately stratified comparisons by time-of-infection and fit multistate models. Results The estimated difference in median LOS for propensity-matched cohorts varied from 3.1 days (95% CI, 2.2–3.9) to 10.1 days (95% CI, 7.3–12.2) depending on the case definition; however, dependency of the estimate on time-to-infection was observed. Stratification by time to first positive toxin assay, excluding probable community-acquired infections, showed a minimum excess LOS of 3.1 days (95% CI, 1.7–4.4). Under the same case definition, the multistate model averaged an excess LOS of 3.3 days (95% CI, 2.6–4.0). Conclusions Two independent time-to-infection adjusted methods converged on similar excess LOS estimates. Changes in LOS can be extrapolated to a marginal dollar costs by multiplying by average costs of an inpatient-day. Infection control officers can leverage automatically extractable EMR data to estimate costs of CDI at their own institution.

show abstract

Approximate Residual Balancing: Debiased Inference of Average Treatment Effects in High Dimensions

Cited by 281 publications

References 85 publications

Double/debiased machine learning for treatment and structural parameters

Double/debiased machine learning for treatment and structural parameters

Estimating bounds on causal effects in high-dimensional and possibly confounded systems

Estimating Local Costs Associated WithClostridium difficileInfection Using Machine Learning and Electronic Medical Records

Contact Info

Product

Resources

About