Algorithms for non-linear Huber estimation (Q1123583): Difference between revisions

The estimation of \textit{P. J. Huber} [Robust statistics (1981; Zbl 0536.62025)] is a modification of least squares estimation designed to reduce the influence of data contamination. Numerical determination of the estimate requires minimization of a sum of \(\rho (f_ j(x))\) where \(\rho\) is the Huber function, and \(f_ j\) are (in general non-linear) regression functions of the parameter vector x. The authors present and investigate algorithms for this task.