On The So-Called “Huber Sandwich Estimator” and “Robust Standard Errors”

By David Freedman. 2006

530 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/e979a1fa0374efe59b7c7cf393ea3639d0db4cd9

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

Modeling Ordered Choices: A Primer and Recent Developments (2008, unknown): “[Additional commentary on this result appears in Freedman (2006).]”
Unbiased Estimation of the Average Treatment Effect in Cluster-Randomized Experiments (2015, methodology): “We refer the reader to Williams (1961), Freedman (2008a) and Freedman (2008b) for additional reading on the particular bias associated with the regression adjustment of random samples and experimental data.”
Discrete Choice Modeling (2007, unknown): “See Freedman (2006).]”
Models as Approximations - A Conspiracy of Random Regressors and Model Deviations Against Classical Inference in Regression (2015, background): “• Weak-sense orthogonalities: η, , δ ⊥ ~ X, that is, (15) E[ ~ X η ] = 0, E[ ~ X ] = 0, E[ ~ X δ ] = 0.”
Trial-by-trial data analysis using computational models (2011, background): “This “Huber-White” or “sandwich” estimator (Huber, 1967; Freedman, 2006) is H − 1 BH − 1 where B = ∑ t g ( c t ) T g ( c t ) , and g ( c t ) , in turn, is the gradient (vector of ﬁrst partial derivatives with respect to the parameters) of the negative log likelihood of the t th data point c t ,…”
The Latent Structure of Spatial Skills and Mathematics: A Replication of the Two-Factor Model (2017, methodology): “0 program (Muthén & Muthén, 1998–2012) using the raw scores for all measures and MLR to guard against specification errors due to non-normal distribution (Freedman, 2006; Muthén & Muthén, 2012; Wang & Wang, 2012).”
Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference (2024, methodology): “To compute confidence intervals on the BT coefficients, we employ two strategies: (1) the pivot bootstrap (DiCiccio & Efron, 1996), and (2) the “sandwich” robust standard errors outlined in Huber et al. (1967) (see also Freedman (2006) for an outline of the necessary technical assumptions).”
Object-oriented Computation of Sandwich Estimators (2006, background): “Some further expository discussion of this issue for ML and quasi ML estimators can be found in Freedman (2006) and Koenker (2006).”
Models as Approximations: How Random Predictors and Model Violations Invalidate Classical Inference in Regression (2014, background): “(A slight subtlety arises from the constraint imposed on η( ~ X) by orthogonality (11) to the predictors, but it does not invalidate the general fact.”
Models as Approximations I: Consequences Illustrated with Linear Regression (2014, methodology): “The answer can be gleaned from Figure 6 which illustrates the proposition for normally distributed Xj•: If nonlinearities and/or heteroskedasticities blow up ... • in the tails of the Xj• distribution, then RAV takes on large values; • in the center of the Xj• distribution, then RAV takes on small v”
Modeling Ordered Choices: A Primer (2010, unknown): “8) and Freedman (2006).] White raises this issue explicitly, although it seems to receive little attention in the literature: “It is the consistency of the QMLE for the parameters of interest in a wide range of situations which insures its usefulness as the basis for robust estimation techniques” (1”
Various Versatile Variances: An Object-Oriented Implementation of Clustered Covariances in R (2020, background): “…in the data. the model’s score function was correctly speciﬁed but that only the remaining likelihood was potentially misspeciﬁed, e.g., due to a lack of independence as in the case of clustered correlations (see White 1994, for a classic textbook, and Freedman 2006, for a criticial review).”
Farmer’s Adaptation to Extreme Weather Events through Farm Management and Its Impacts on the Mean and Risk of Rice Yield in China (2015, unknown): “This approach yields consistent estimates of the covariance matrix without making distributional assumptions (Freedman 2006).”