Regression‐assisted inference for the average treatment effect in paired experiments

By C. Fogarty. Biometrika, 2016

62 citations recorded by the source corpus.

Summary In paired randomized experiments, individuals in a given matched pair may differ on prognostically important covariates despite the best efforts of practitioners. We examine the use of regression adjustment to correct for persistent covariate imbalances after randomization, and present two regression‐assisted estimators for the sample average treatment effect in paired experiments. Using the potential outcomes framework, we prove that these estimators are consistent for the sample average treatment effect under mild regularity conditions even if the regression model is improperly specified, and describe how asymptotically conservative confidence intervals can be constructed. We demonstrate that the variances of the regression‐assisted estimators are no larger than that of the standard difference‐in‐means estimator asymptotically, and illustrate the proposed methods by simulation. The analysis does not require a superpopulation model, a constant treatment effect, or the truth of the regression model, and hence provides inference for the sample average treatment effect with the potential to increase power without unrealistic assumptions.

Subfields: Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/ebebb556d9fa034d6dcd020f91904ec306ebaeba

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

1. Kevin Guo
2. Guillaume W. Basse
3. Peng Ding
4. Xinran Li
5. Anqi Zhao
6. Hanzhong Liu
7. Yuehan Yang
8. Jason A. Wu

Citations in the corpus

• Decomposing Treatment Effect Variation (2016, unknown): “See also Fogarty (2016).”
• Regression-adjusted average treatment effect estimates in stratified randomized experiments (2019, background): “Fogarty [9, 10] discussed the advantages of regression adjustment in two special stratified randomized experiments, finely stratified and paired experiments.”
• Rerandomization and regression adjustment (2019, background): “In general, these confidence intervals can be conservative and can be improved (Aronow et al., 2014; Fogarty, 2018; Ding et al., 2019).”
• The Generalized Oaxaca-Blinder Estimator (2020, background): “Theorem 1 has since been generalized to other experimental designs (Fogarty 2018; Li and Ding 2020; Liu and Yang 2020) and to high-dimensional linear regression (Bloniarz et al. 2016; Lei and Ding 2020).”
• Randomization Tests for Weak Null Hypotheses in Randomized Experiments (2018, background): “Our theory does not encapsulate SREs with many small strata, that is, the N[h]’s are bounded but H → ∞ (Fogarty 2018a).”
• Covariate-adjusted Fisher randomization tests for the average treatment effect (2020, methodology): “We conjecture that the strategy of appropriately studentizing an efficient, covariate-adjusted estimator works for frt in general experiments as well (e.g., Dasgupta et al. 2015; Lu 2016; Mukerjee et al. 2018; Middleton 2018; Fogarty 2018a,b).”