Summary

In almost any area of scientific research there is an increasing availability of data. These data sets are not only becoming larger in size, but also in complexity.  Extracting essential features and finding structures and relations in data sets can be done with statistical methods.  But it was observed that many classical statistical techniques are not resistant in presence of outliers and become completely invalid. We need therefore to develop robust statistical procedures that give reliable results also when atypical observations are present, a problem that occurs frequently when analysing complex data sets.  Moreover, there is also a need for diagnostic tools to pinpoint these outliers since they may contain valuable information. While robust methods are well established for studying data sets with simple models, this no longer holds for more complicated, multivariate and non-linear models. This Network will reinforce research efforts for analysing such complex data sets, not only with robust methods, but also with non-parametric and other ones.

Duration

3 years, from January 2004 to December 2006.