A Variant of Huber Robust Regression
In this paper the authors present a new variant of Huber's classic M- estimator for robust regression dealing with the scale estimate differently than Huber. The class of estimators developed includes least squares (good for small errors) and least absolute residuals (good for large errors) as limiting cases and can be considered as a generalization of the trimmed mean to regression. In addition, for asymmetric contamination the estimator proposed should perform better than the trimmed mean. The authors exhibit a so-called ABC algorithm to compute this estimator and to prove its correctness. Also, they show how to extend their estimators to include weighting, equality and inequality constraints, and the addition or deletion of data points.
- scientific article; zbMATH DE number 3905646
- The scale problem in robust regressionM- estimates
- Robust regression function estimation
- On Newton's method for Huber's robust M-estimation problems in linear regression
- High Breakdown-Point Estimates of Regression by Means of the Minimization of an Efficient Scale
- Parallel proximal decomposition algorithms for robust estimation.
- Robust regression through the Huber's criterion and adaptive lasso penalty
- Robust regression trees based on M-estimators
- A New Scale Step for Huber’s M-Estimators in Multiple Regression
- A variational approach to robust regression
- scientific article; zbMATH DE number 7730941 (Why is no real title available?)
- Another look at Huber's estimator: a new minimax estimator in regression with stochastically bounded noise
This page was built for publication: A Variant of Huber Robust Regression
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3334793)