Estimators under a semi-functional linear regression model

VENA, PABLO; SALIBIAN-BARRERA, MATÍAS; BOENTE, GRACIELA

Leuven

Conferencia; International Conference on Robust Statistics (ICORS 2018); 2019

KU Leuven

In this talk, we propose robust estimators for the regression coefficient and the non-parametric component under a semi-functional linear regression model. The interest is in reducing the damaging eect on the estimator caused by potentially atypical observations both in the response variable and in the functional covariate. For that reason, we consider a robust alternative for principal directions using spherical principal components and we use MM􀀀estimators combined with B-splines for the nonparametric component. The amount of regularization for the functional regression coefficient and the non-parametric component are chosen using a robust BIC-type criterion.