Large Dimensional Analysis of Robust M-Estimators of Covariance with Outliers
arXiv:1503.01245 · doi:10.1109/TSP.2015.2460225
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.
arXiv commentSubmitted to IEEE Transactions on Signal Processing
References in corpus (8)
- Regularized estimation of large covariance matrices
- Large System Analysis of Linear Precoding in Correlated MISO Broadcast Channels under Limited Feedback
- Robust Shrinkage Estimation of High-dimensional Covariance Matrices
- The spectrum of kernel random matrices
- A Robust Statistics Approach to Minimum Variance Portfolio Optimization
- Robust Estimates of Covariance Matrices in the Large Dimensional Regime
- Asymptotic Moments for Interference Mitigation in Correlated Fading Channels
- On the convergence of Maronna's -estimators of scatter
Cited by in corpus (8)
- A Robust Statistics Approach to Minimum Variance Portfolio Optimization
- Efficient Computation of Limit Spectra of Sample Covariance Matrices
- Robust Sparse Covariance Estimation by Thresholding Tyler's M-Estimator
- Analysis and Design of Multiple-Antenna Cognitive Radios with Multiple Primary User Signals
- Large-dimensional behavior of regularized Maronna's M-estimators of covariance matrices
- Marchenko-Pastur Law for Tyler's M-estimator
- Tyler's and Maronna's M-estimators: Non-Asymptotic Concentration Results
- On eigenvalues of a high-dimensional spatial-sign covariance matrix