Segmented regression: a robust approach

Abstract

Robust estimators are developed for the segmented regression model, a model consisting of two linear segments separated by a change-point. Julious (2001) introduced a method to estimate parameters in the case of unknown changepoint. The focus of this practicum is on robustifying the Julious algorithm via iteratively re-weighting, extending the work of Julious (2001). Simulation studies are conducted to assess the performance of the iterative re-weighting. The methods are applied to a physiological data set studied by Julious, and to two stock-recruit data sets from fisheries science .