Comment: Performance of Double-Robust Estimators When “Inverse Probability” Weights Are Highly Variable

By James M. Robins, Mariela Sued, Q. Lei-Gomez, Andrea Rotnitzky. 2007

455 citations recorded by the source corpus.

Comment on ``Performance of Double-Robust Estimators When ``Inverse Probability'' Weights Are Highly Variable'' [arXiv:0804.2958]

Subfields: Statistics and Probability

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

Double/De-Biased Machine Learning of Global and Local Parameters Using Regularized Riesz Representers (2018, background): “The earliest reference of which we know is Robins et al. (2007), a comment on another paper, which consider only the global average treatment effect (ATE).”
Intrinsically Efficient, Stable, and Bounded Off-Policy Evaluation for Reinforcement Learning (2019, methodology): “Replacing, ω0:t with its self-normalized version ω0:t/En [ω0:t] in (2) leads to SNDR [18, 26] (aka WDR), but it only satisfies these properties partially: it’s only 2-bounded and partially stable (see Appendix A).”
Multiply Robust Estimation in Regression Analysis With Missing Data (2014, background): “There have been some recent developments trying to overcome this difficulty, including Kang and Schafer (2007), Robins et al. (2007), and Cao, Tsiatis, and Davidian (2009).”
Sensitivity analysis for inverse probability weighting estimators via the percentile bootstrap (2017, methodology): “Compared to IPW, the SIPW estimator is sample bounded (Robins et al., 2007), that is,”
Doubly robust difference-in-differences estimators (2018, background): “An important part of these potential instabilities comes from the fact that the weights associated with Horvitz-Thompson type estimators do not sum up to one, see e.g. Robins et al. (2007).”
Minimal dispersion approximately balancing weights: asymptotic properties and practical considerations (2017, background): “This is related to the sample boundedness property discussed in Robins et al. (2007), which requires the estimator to lie within the range of observed values of the outcome.”
New Evidence on the Finite Sample Properties of Propensity Score Reweighting and Matching Estimators (2014, methodology): “The normalized reweighting estimator is discussed in Lunceford and Davidian (2004), Imbens (2004), and Robins et al. (2007), for example.”
NICE DSU Technical Support Document 18: Methods for population-adjusted indirect comparisons in submissions to NICE (2016, unknown): “criteria (MAIC ESS): For C1: 35 (29) For C2: 35 (15) For C3: 57 (14) For C4: 35 (26) For C5: 19 (17) C1: 30 C2: 18 C3: 95 C4: 40 C5: 109 5-6 0”
An Introduction to the Augmented Inverse Propensity Weighted Estimator (2010, methodology): “The key advantage of the AIPW estimator (and other more advanced doublerobust estimators [Robins et al. 2007]) in such a situation is that it will continue to perform well as long as either the propensity score model or the outcome regressions are properly specified.”
A Survey on Causal Inference (2020, background): “ly explain the relationship among variables. The combination of outcome regression with weighting by propensity score ensures that the estimators are robust to misspecification of one of these models [12, 101, 103, 117]. The DR estimator consults outcomes to make the IPW estimator robust when propen”
Augmented minimax linear estimation (2017, methodology): “This phenomenon is not unique to our approach, as some other methods can estimate something like a Riesz representer on a tangent space of their choosing; see, e.g., Remark 2.5 of Chernozhukov et al. (2018a) or Section 3 of Robins et al. (2007).”
A GENERAL DOUBLE ROBUSTNESS RESULT FOR ESTIMATING AVERAGE TREATMENT EFFECTS (2014, unknown): “A similar result had also been demonstrated by Robins et al. (2007).”
Combining Inverse Probability Weighting and Multiple Imputation to Improve Robustness of Estimation (2016, background): “In such a case, a more feasible way, as discussed in Robins et al. (2007), is to fit multiple parametric models for .”