B1646
Title: Robust techniques to estimate linear regression models for interval-valued data
Authors: Angela Blanco-Fernandez - University of Oviedo (Spain) [presenting]
Gil Gonzalez-Rodriguez - University of Oviedo (Spain)
Abstract: Some robust statistical techniques to estimate linear regression models for interval-valued data are shown. The least-squares (LS) estimation of the models has been previously solved through constrained optimization techniques guaranteeing the coherence of the solutions within the interval framework. However, as happens with classical estimation techniques of regression problems, the LS estimators are not robust againts outliers. Thus, alternative estimation approaches should be developed. One of the best-known procedures in classical regression is the least trimmed estimator. It is based on estimating the regression coefficients by trimming the sample dataset, discarding the observations with greater residuals. The technique is applied to the interval framework. Several alternatives are investigated, and some applications and simulations are shown to illustrate the behaviour of the procedures.
