What question did this study set out to answer?

To develop a computationally tractable and robust method for estimating local average treatment effects with instrumental variables in high-dimensional covariate settings.

August 25, 2021Journal of Business and Economic Statistics

High-Dimensional Model-Assisted Inference for Local Average Treatment Effects With Instrumental Variables

View Full Paper

Ask AI

Bookmark

Authors

BSBaoluo SunHarvard University ZTZhiqiang TanFirst Affiliated Hospital of Jinan University

Discussion

Member takes

Implication

Methodological analysis demonstrates valid treatment effect inference with high-dimensional instrumental variables, indicating robust estimation even under outcome model misspecification.

Key Points

To develop a computationally tractable and robust method for estimating local average treatment effects with instrumental variables in high-dimensional covariate settings.
Combined regularized calibrated estimation for regression parameters with a doubly robust point estimator for treatment parameters.
Derived theoretical conditions for Wald confidence interval validity under high-dimensional sparsity when nuisance models are partially misspecified.
Assessed finite-sample performance through numerical simulations and an empirical application estimating the returns to education, implemented in the R package RCAL.
Established theoretically that Wald confidence intervals remain asymptotically valid if the instrument propensity score is correctly specified, even when treatment and outcome regression models are misspecified.
Eliminated the strict requirement of standard high-dimensional estimators that demand correct specification across all three nuisance models simultaneously.

Cite This Study

Sun et al. (2021) studied this question.

synapsesocial.com/papers/6a7a2a4414b028facf0e32a5 https://doi.org/10.1080/07350015.2021.1970575

View Full Paper

Ask AI

Bookmark

Implication

Methodological analysis demonstrates valid treatment effect inference with high-dimensional instrumental variables, indicating robust estimation even under outcome model misspecification.

High-Dimensional Model-Assisted Inference for Local Average Treatment Effects With Instrumental Variables

Discussion

Member takes

Implication

Key Points

Cite This Study

High-Dimensional Model-Assisted Inference for Local Average Treatment Effects With Instrumental Variables

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider