Robust Discrimination under a Hierarchy on the Scatter Matrices

ANA BIANCO, GRACIELA BOENTE, ANA PIRES E ISABEL RODRIGUES

Lisboa, Portugal

Congreso; International Conference on Robust Statistics (ICORS 2006); 2006

We study the properties of robust quadratic discrimination rules based on robust estimates of the involved parameters. Our analysis is based on the partial influence functions (Pires and Branco, 2002) of these functionals, related with the robust estimators coefficients and we derive the partial influence functions of the missclasification probabilities, using the results in Croux and Joossens (2005). Through them , we obtain the asymptotic variances of the estimated coefficients under models (i) to (iv) and we compare them across all correct models in a given situation. This comparison leads us to the same conclusions as those of in Flury and Schmid (1992).