Robust procedures for single index models with asymmetric errors

BOENTE, GRACIELA; BIANCO, ANA; AGOSTINELLI, CLAUDIO

Sevilla

Conferencia; IX International Conference of the ERCIM Working Group on Computing and Statistics (ERCIM2016); 2016

Universidad de Sevilla, Queen Mary University, Birbeck University of London

family of robust estimators is introduced for the link function and the singleindex parameter under a single index model with asymmetric errors. The proposal is based on a three-step procedure related to a profile approach that combines parametric and nonparametric techniques. The procedure controls high deviance or Pearson residuals through a bounded loss function. Its asymptotic behaviour is derived under mild assumptions. A numerical study illustrates the finite sample behaviour of the proposed procedures in different scenarios.