One-step GM-estimators for orthogonal refression

Abstract

This paper presents a one–step robust generalised M-estimation for orthogonal regression. The GM-estimator uses Schweppe weights which are based on high breakdown initial and scale estimates to downweight outliers and high leverage points. The one-step iteratively reweighted least squares procedure was used to compute the GM estimates. The robustness of the GM-estimator was shown from the results illustrated on measurements of concrete compressive strengths data.

Keywords:

• Functional errors-in-variables model
• Orthogonal regression
• GM-estimators
• bootstrap method