Co-chairs: Stefan Van Aelst and Elvezio Ronchetti
Description: This WG will collaborate with WG A for the development of Task 1, and with WG C for the development of Task 3. A sound framework will be established to handle non-standard and non-perfect datasets, including elements such as functions, surfaces or sets, e.g. those provided by WG A. New robust population measures and models will be investigated, since the location-scale model is not always meaningful, as it relies on a Euclidean structure. Important problems which have not previously been considered will be clearly identified, formalized and tackled. Problems concerning location, including different regression frameworks, scale, clustering, and dimensionality reduction involving non-standard data, will be considered as starting points. Robust methods for estimation, such as those based on trimming, are intuitively generalizable. Alternatives using approaches such as M-type and minimum divergence estimators or distance- based methods will be tackled. Robust filtering techniques will be developed.