The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials

By Donald B. Rubin. 2007

1,145 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, methodology): “In part of the discussion in this chapter, therefore, we view the sample at hand as the full population of interest, following the approaches taken by Fisher (1925, 1935), Neyman (1935), and subsequently by Rubin (1974, 1978, 2007).”
Causal inference in statistics: An overview (2009, neutral): “It should be emphasized, however, that contrary to conventional wisdom (e.g., (Rubin, 2007, 2009)), propensity score methods are merely efficient estimators of the right hand side of (25); they cannot be expected to reduce bias in case the set S does not satisfy the back-door criterion (Pearl,…”
Homelessness and Reentry (2014, background): “…a nonissue because propensity score matching is not attempting to assess the unique contribution of each of the variables included in the model, but the match provided by all predictors collectively, that is, specifying prediction rather than specifying causation (Rubin, 2006, 2007; Stuart, 2010).”
For objective causal inference, design trumps analysis (2008, background): “The same point has been emphasized in Rubin (2002, 2007) and the subsequent editorial by D’Agostino and D’Agostino (2007).”
Assessing Treatment Effect Variation in Observational Studies: Results from a Data Challenge (2019, background): “, Stuart, E. A., Hill, J. L., and Green, D. P. (2016). Assessing methods for generalizing experimental impact estimates to target samples. Journal of Research in Educational Effectiveness, 9:103–127. Murray, J. S. (2017). Log-Linear Bayesian Additive Regression Trees for Categorical and Count Respon”
Covariate balancing propensity score (2014, methodology): “Xi/}]−1X̃iX̃Ti ) .16/ when we use the average treatment effect weight given in equation (10), or Σβ.”
Causal Effects of Perceived Immutable Characteristics (2011, background): “For the remainder of this paper, we assume familiarity with the following concepts: units, treatments, timing of treatment assignment, outcome variables, covariates (observed and unobserved), intermediate outcomes, assignment mechanism, ignorability, posttreatment-adjustment bias, the stable unit tr”
An introduction to propensity-score methods for reducing confounding in observational studies (2011, positive): “Second, propensity score methods allow one to separate the design of an observational study from its analysis (Rubin, 2007).”
Globally efficient non-parametric inference of average treatment effects by empirical balancing calibration weighting (2015, methodology): “For balancing an overidentified system of equations by using the empirical likelihood, there is no guarantee that the γ-estimate is √ N consistent (Schennach, 2007) under a misspecified propensity score model.”
A new criterion for confounder selection (2011, background): “A recent series of letters and exchanges between Shrier (2008), Rubin (2007, 2008, 2009), Pearl (2009), and Sjölander (2009) raised the question as to whether one should always adjust, either by regression or propensity score analysis (Rosenbaum and Rubin, 1983), for all pretreatment covariates.”
Estimating Individual Treatment Effect in Observational Data Using Random Forest Methods (2017, background): “In recent innovations, these RF approaches have been used specifically to uncover subgroups with differential treatment responses (Su et al. 2009, 2011; Foster, Taylor, and Ruberg 2011).”
Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies (2015, methodology): “We suggest that statistical hypothesis testing in the analytic sample not be used to identify the requisite variables, in the spirit of separating ‘design’ from ‘analysis’ and not using the outcome data in the propensity score process [22].”