On regression adjustments to experimental data

By David Freedman. 2008

478 citations recorded by the source corpus.

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

Citations in the corpus

The Econometrics of Randomized Experiments (2016, background): “Since randomization does not justify the models, almost anything can happen” (Freedman, 2008, abstract) and similar comments are made by Deaton (2010) Young (2016), and Imbens and Rubin (2015).”
Revisiting regression adjustment in experiments with heterogeneous treatment effects (2020, background): “The efficiency argument in favor of regression adjustment has come under criticism lately, where papers like Freedman (2008a,b) find no systematic gain in asymptotic efficiency of the covariate adjusted estimator. In this paper, we verify that when treatment effects are heterogeneous, additively con”
Lasso adjustments of treatment effect estimates in randomized experiments (2015, methodology): “Remark 4: If we assume δn =Oð ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi log p=n p Þ, which satisfies [10], then Condition 6 requires that the tuning parameters are proportional to ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi log p=n p , which is typically assumed for the Lasso in the high-dim”
Machine Learning and Causal Inference : A Modular Approach to Assessing the Effects of the London Bombings of 2005 (2015, background): “…to two equal sized groups or the experiment is small, some might point out that conventional covariance adjustment (i.e. adding a covariate like a control variable to a linear model) produces biased estimates of treatment effects (and biased statistical tests) (Freedman, 2008; Lin et al., 2013).”
Finite Population Causal Standard Errors (2014, methodology): “More recent studies analyze regression methods with additional covariates under the randomization distribution in randomized experiments, e.g., Freedman (2008ab), Lin (2013), and Schochet (2010).”
Understanding and misunderstanding randomized controlled trials (2016, unknown): “Freedman (2008) shows that the adjusted estimate of the ATE is biased in finite samples, with the bias depending on the correlation between the squared treatment effect and the covariates.”
Rerandomization to improve covariate balance in experiments (2012, unknown): “ve sources. If covariates are not balanced before the experiment, post-hoc methods such asregression adjustment are commonly used,which rely on assumptions that often cannot be veriﬁed [Tukey (1993); Freedman (2008)]. Moreover, unlike post-hoc methods,rerandomization isconducted entirely at the desi”
Better Late than Nothing: Some Comments on Deaton (2009) and Heckman and Urzua (2009) (2010, methodology): “In the absence of interactions we can estimate the average effect without bias, but the validity of confidence intervals still relies on large sample approximations (e.g., Neyman, 1923; Freedman, 2008).”
General Forms of Finite Population Central Limit Theorems with Applications to Causal Inference (2016, methodology): “Although the simple difference-in-means estimator of the average causal effect is unbiased, carefully using the covariates often improves the efficiency (Freedman 2008a, b; Lin 2013).”
Adjusting treatment effect estimates by post‐stratification in randomized experiments (2013, methodology): “Freedman (2008a,b) analysed the regression-adjusted estimator under the Neyman– Rubin model without treatment-by-strata interactions and found that the asymptotic variance might be larger than if no correction were made. Lin (2012) extended Freedman’s results and showed that, when a treatment-by-cov”
Opiates for the Matches: Matching Methods for Causal Inference (2009, background): “The regression standard errors can be severely biased, and the multiple regression estimator may have higher asymptotic variance than simply estimating Equation 3 (for details see Freedman 2008b,c).”
Statistical Power for Regression Discontinuity Designs in Education Evaluations (2009, methodology): “These assumptions are consistent with ordinary least squares (OLS) methods that are typically used to estimate program impacts in education research, and are required to ensure that variances based on OLS methods are justified by the Neyman model of causal inference (Freedman, 2008; Schochet, 2007).”