Statistical Models and Causal Inference: Randomization Does Not Justify Logistic Regression

By David Freedman, David Collier, Jasjeet S. Sekhon, Philip B. Stark. 2009

168 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/3e685939f389faf6c94dd29d04f86d5a7051aef9

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 Generalized Oaxaca-Blinder Estimator (2020, methodology): “Like the results of prior work on covariate adjustment with linear models (Bloniarz et al. 2016; Fogarty 2018; Freedman 2008a,b; Li and Ding 2020; Liu and Yang 2020), our theoretical guarantees are asymptotic.”
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).”
On regression adjustments in experiments with several treatments (2008, background): “For additional discussion, see Freedman (2007). An example may clarify the issues.”
The International Journal of Biostatistics Empirical Efficiency Maximization : Improved Locally Efficient Covariate Adjustment in Randomized Experiments and Survival Analysis (2011, unknown): “Robinson and Jewell (1991) demonstrate asymptotic efficiency costs for naive adjustment, and Freedman (2008) provides more commentary.”
A Class of Unbiased Estimators of the Average Treatment Effect in Randomized Experiments (2013, background): “Freedman [6, 7, 36] rekindled interest in the design-based analysis of randomized experiments.”
Randomization in the Design of Experiments (2009, background): “Freedman (2008) has discussed the non-null randomization theory of logistic regression, emphasizing that the standard logistic regression coefficient does not have a direct interpretation in terms of the following extension of equation (1).”
Regression-adjusted average treatment effect estimates in stratified randomized experiments (2019, background): “It would be interesting to extend the results to more complicated settings, such as covariates adjustment for multiple value treatment [11], for binary outcomes using logistic regression [12, 24, 35], and covariate adjustment in experiments when there is noncompliance [2, 3, 17].”
Extension of the modified Poisson regression model to prospective studies with correlated binary data (2013, background): “That is, the estimates of odds ratio obtained from the logistic regression model are not a consistent estimate of the average effect.(6) In contrast, risk ratio estimates are free of such”
Bridging Finite and Super Population Causal Inference (2017, background): “not new, we oﬀer it as it gives an alternative derivation that does not rely on asymptotics such as a growing superpopulation or a focus on the properties of the treatment assignment mechanism. Using Freedman (2008a)’s results, Aronow et al. (2014) considered a super population with N units, with th”
Lasso adjustments of treatment effect estimates in randomized experiments (2015, methodology): “The asymptotic properties of this kind of estimator are explored under the Neyman-Rubin model in [11], [12], and [7].”
A paradox from randomization-based causal inference (2014, methodology): “Although linear regression has been criticized for analyzing experimental data (Freedman, 2008), the least square estimator β̂OLS = τ̂ is unbiased for the average causal effect τ .”