Influence Functions of Iteratively Reweighted Least Squares Estimators

From MaRDI portal





The authors study the influence functions of iteratively reweighted least squares estimators. The influence function of the weighted least squares estimator is shown to depend both on the weights and on the influence function of the initial estimator, where the weights are determined by both position in the design space and the residual relative to an initial regression estimator. When the weights are iteratively reestimated, convergence of the corresponding estimates and their influence functions relies on the magnitude of the iteration derivative, which is the derivative of the \(k+1\)st iterate with respect to the \(k\) th iterate. Conditions are given on the weights for such convergence and for boundedness of the limiting influence function, and shown to be easily checked for the Mallows class of estimators, including Huber's \(M\) estimators. Also, the relationship between iteratively reweighted least squares, Newton-Raphson one-step estimators, and the fully iterated \(M\) estimator they approximate is clarified. Besides, a numerical study for Huber's \(M\) estimators has been carried out to show that the iteration derivative determines not only the rate of convergence of the sequence of influence functions corresponding to iterated reweighted least squares but the rate of convergence of the iterated estimates themselves. Further, the case with unknown scale is discussed.











This page was built for publication: Influence Functions of Iteratively Reweighted Least Squares Estimators

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3984604)