Finite algorithms for robust linear regression (Q2638753)

Two algorithms for solving the linear estimation problem of \textit{P. J. Huber} [Ann. Math. Stat. 35, 73-101 (1964; Zbl 0136.398)] in a linear model are defined and their properties are studied. It is shown that both algorithms are efficient and in the normal, i.e. non degenerate, case there is no significant difference between their respective performance as measured by the number of iterations or by the computing time. Both the results of the simulation experiments with a large number of test problems and the theoretical conclusions show that both algorithms can be efficiently used for computing the \(\ell_ 1\)-norm solution.