Covariate selection with group lasso and doubly robust estimation of causal effects

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

The efficiency of doubly robust estimators of the average causal effect (ACE) of a treatment can be improved by including in the treatment and outcome models only those covariates which are related to both treatment and outcome (i.e., confounders) or related only to the outcome. However, it is often challenging to identify such covariates among the large number that may be measured in a given study. In this article, we propose GLiDeR (Group Lasso and Doubly Robust Estimation), a novel variable selection technique for identifying confounders and predictors of outcome using an adaptive group lasso approach that simultaneously performs coefficient selection, regularization, and estimation across the treatment and outcome models. The selected variables and corresponding coefficient estimates are used in a standard doubly robust ACE estimator. We provide asymptotic results showing that, for a broad class of data generating mechanisms, GLiDeR yields a consistent estimator of the ACE when either the outcome or treatment model is correctly specified. A comprehensive simulation study shows that GLiDeR is more efficient than doubly robust methods using standard variable selection techniques and has substantial computational advantages over a recently proposed doubly robust Bayesian model averaging method. We illustrate our method by estimating the causal treatment effect of bilateral versus single-lung transplant on forced expiratory volume in one year after transplant using an observational registry.

Original languageEnglish (US)
Pages (from-to)8-17
Number of pages10
JournalBiometrics
Volume74
Issue number1
DOIs
StatePublished - Mar 2018

Fingerprint

Causal Effect
Lasso
Robust Estimation
Covariates
Transplants
Variable Selection
Forced Expiratory Volume
methodology
Bayesian Model Averaging
Registries
Coefficient Estimates
Robust Estimators
Averaging Method
Consistent Estimator
Robust Methods
Treatment Effects
Lung
lungs
alachlor
Predictors

Keywords

  • Average treatment effect
  • Causal inference
  • Group lasso
  • Variable selection

Cite this

Covariate selection with group lasso and doubly robust estimation of causal effects. / Koch, Brandon; Vock, David M.; Wolfson, Julian.

In: Biometrics, Vol. 74, No. 1, 03.2018, p. 8-17.

Research output: Contribution to journalArticle

@article{911ad541f714464381450cc2a325b077,
title = "Covariate selection with group lasso and doubly robust estimation of causal effects",
abstract = "The efficiency of doubly robust estimators of the average causal effect (ACE) of a treatment can be improved by including in the treatment and outcome models only those covariates which are related to both treatment and outcome (i.e., confounders) or related only to the outcome. However, it is often challenging to identify such covariates among the large number that may be measured in a given study. In this article, we propose GLiDeR (Group Lasso and Doubly Robust Estimation), a novel variable selection technique for identifying confounders and predictors of outcome using an adaptive group lasso approach that simultaneously performs coefficient selection, regularization, and estimation across the treatment and outcome models. The selected variables and corresponding coefficient estimates are used in a standard doubly robust ACE estimator. We provide asymptotic results showing that, for a broad class of data generating mechanisms, GLiDeR yields a consistent estimator of the ACE when either the outcome or treatment model is correctly specified. A comprehensive simulation study shows that GLiDeR is more efficient than doubly robust methods using standard variable selection techniques and has substantial computational advantages over a recently proposed doubly robust Bayesian model averaging method. We illustrate our method by estimating the causal treatment effect of bilateral versus single-lung transplant on forced expiratory volume in one year after transplant using an observational registry.",
keywords = "Average treatment effect, Causal inference, Group lasso, Variable selection",
author = "Brandon Koch and Vock, {David M.} and Julian Wolfson",
year = "2018",
month = "3",
doi = "10.1111/biom.12736",
language = "English (US)",
volume = "74",
pages = "8--17",
journal = "Biometrics",
issn = "0006-341X",
publisher = "Wiley-Blackwell",
number = "1",

}

TY - JOUR

T1 - Covariate selection with group lasso and doubly robust estimation of causal effects

AU - Koch, Brandon

AU - Vock, David M.

AU - Wolfson, Julian

PY - 2018/3

Y1 - 2018/3

N2 - The efficiency of doubly robust estimators of the average causal effect (ACE) of a treatment can be improved by including in the treatment and outcome models only those covariates which are related to both treatment and outcome (i.e., confounders) or related only to the outcome. However, it is often challenging to identify such covariates among the large number that may be measured in a given study. In this article, we propose GLiDeR (Group Lasso and Doubly Robust Estimation), a novel variable selection technique for identifying confounders and predictors of outcome using an adaptive group lasso approach that simultaneously performs coefficient selection, regularization, and estimation across the treatment and outcome models. The selected variables and corresponding coefficient estimates are used in a standard doubly robust ACE estimator. We provide asymptotic results showing that, for a broad class of data generating mechanisms, GLiDeR yields a consistent estimator of the ACE when either the outcome or treatment model is correctly specified. A comprehensive simulation study shows that GLiDeR is more efficient than doubly robust methods using standard variable selection techniques and has substantial computational advantages over a recently proposed doubly robust Bayesian model averaging method. We illustrate our method by estimating the causal treatment effect of bilateral versus single-lung transplant on forced expiratory volume in one year after transplant using an observational registry.

AB - The efficiency of doubly robust estimators of the average causal effect (ACE) of a treatment can be improved by including in the treatment and outcome models only those covariates which are related to both treatment and outcome (i.e., confounders) or related only to the outcome. However, it is often challenging to identify such covariates among the large number that may be measured in a given study. In this article, we propose GLiDeR (Group Lasso and Doubly Robust Estimation), a novel variable selection technique for identifying confounders and predictors of outcome using an adaptive group lasso approach that simultaneously performs coefficient selection, regularization, and estimation across the treatment and outcome models. The selected variables and corresponding coefficient estimates are used in a standard doubly robust ACE estimator. We provide asymptotic results showing that, for a broad class of data generating mechanisms, GLiDeR yields a consistent estimator of the ACE when either the outcome or treatment model is correctly specified. A comprehensive simulation study shows that GLiDeR is more efficient than doubly robust methods using standard variable selection techniques and has substantial computational advantages over a recently proposed doubly robust Bayesian model averaging method. We illustrate our method by estimating the causal treatment effect of bilateral versus single-lung transplant on forced expiratory volume in one year after transplant using an observational registry.

KW - Average treatment effect

KW - Causal inference

KW - Group lasso

KW - Variable selection

UR - http://www.scopus.com/inward/record.url?scp=85021192562&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85021192562&partnerID=8YFLogxK

U2 - 10.1111/biom.12736

DO - 10.1111/biom.12736

M3 - Article

C2 - 28636276

AN - SCOPUS:85021192562

VL - 74

SP - 8

EP - 17

JO - Biometrics

JF - Biometrics

SN - 0006-341X

IS - 1

ER -