Robuste schätzungen: Ein anwendungsorientierter Überblick

The paper offers an introductory survey of the area of robust estimation for applied statisticians, without demanding much mathematics. At first, the most important basic ideas of the robustness theory are explained using an example from the analysis of variance. Then the general reasons for and limits of the “robustification” of statistics are delineated. In particular, the various types of deviations from ideal models, such as gross errors, are discussed and their consequences are pointed out. The next part describes several simple robust estimators and their properties, ending with an outlook on robust estimators in linear models. As an appendix a number of frequent objections to and misunderstandings of robustness theory are discussed, refuted and clarified.