B1121
Title: Distance-based regression using robust Gower's distance
Authors: Eva Boj - Universidad Carlos III de Madrid (Spain) [presenting]
Aurea Grane Chavez - Universidad Carlos III de Madrid (Spain)
Abstract: A robust version of Gower's distance is proposed to be used in the predictors' space of distance-based predictive models. Models under evaluation are the distance-based generalized linear models, which can be used for classification purposes. The performance of the new proposal is compared to that of classical Gower's metric in the presence of outliers in data sets of multivariate heterogeneous data. Mean squared error and other goodness of fit measures are used to evaluate the effectiveness in predicting responses. Computations on real data sets are made using the dbstats package for R.
