Abstract:A large dimensional characterization of robust M-estimators of covariance (or scatter) is provided under the assumption that the dataset comprises independent (essentially Gaussian) legitimate samples as well as arbitrary deterministic samples, referred to as outliers. Building upon recent random matrix advances in the area of robust statistics, we specifically show that the so-called Maronna M-estimator of scatter asymptotically behaves similar to well-known random matrices when the population and sample sizes grow together to infinity. The introduction of outliers leads the robust estimator to behave asymptotically as the weighted sum of the sample outer products, with a constant weight for all legitimate samples and different weights for the outliers. A fine analysis of this structure reveals importantly that the propensity of the M-estimator to attenuate (or enhance) the impact of outliers is mostly dictated by the alignment of the outliers with the inverse population covariance matrix of the legitimate samples. Thus, robust M-estimators can bring substantial benefits over more simplistic estimators such as the per-sample normalized version of the sample covariance matrix, which is not capable of differentiating the outlying samples. The analysis shows that, within the class of Maronna's estimators of scatter, the Huber estimator is most favorable for rejecting outliers. On the contrary, estimators more similar to Tyler's scale invariant estimator (often preferred in the literature) run the risk of inadvertently enhancing some outliers.

Statistical Inference for a Robust Measure of Multiple Correlation

Asymptotic confidence interval for R2 in multiple linear regression

Inferences about robust measures of effect size for two dependent variables including situations where there is a covariate

A NEW PERSPECTIVE ON ROBUST M-ESTIMATION: FINITE SAMPLE THEORY AND APPLICATIONS TO DEPENDENCE-ADJUSTED MULTIPLE TESTING

Is Distance Correlation Robust?

Heteroskedasticity-robust inference in linear regression models with many covariates

Statistical Inference and Large-scale Multiple Testing for High-dimensional Regression Models

Reprint of: Robust inference on correlation under general heterogeneity

Statistical inference based on robust low-rank data matrix approximation

Robust Estimation of Polychoric Correlation

Bayesian Robustness to Outliers in Linear Regression and Ratio Estimation

Robust Inference Under Heteroskedasticity via the Hadamard Estimator

Robust Inference under the Beta Regression Model with Application to Health Care Studies

Statistical Distances and Their Role in Robustness

Universal Robust Regression via Maximum Mean Discrepancy

Large Dimensional Analysis of Robust M-Estimators of Covariance with Outliers

With random regressors, least squares inference is robust to correlated errors with unknown correlation structure

Robust estimation and test for Pearson's correlation coefficient

Robust Inference for Seemingly Unrelated Regression Models

On the minimax robustness against correlation and heteroscedasticity of ordinary least squares among generalized least squares estimators of regression