Robust statistics - an Introduction 

			     Jana Noskova

Classical statistical methods of estimation (e.g. MLE ) are usually
very sensitive to deviations from the assumed idealized model.  In
response, robust methods of estimation, i.e.  methods are not so
sensitive to small deviations from an underlying model, have been
developed.

We will review the parametric concept of estimation and then introduce
the robust approach. Some basic robust estimators of location
(e.g. trimmed mean, Hampel`s estimator) and measures of robustness
(influence function, breakdown point) will be described.