(University of Wuerzburg)
In statistics, a record is an observation that is larger than its predecessors. However, if the observations are not real numbers but vectors or even continuous functions, it is not obvious how to define a record anymore. Different concepts of multivariate records are proposed, and their stochastical behaviour is investigated. The key assumption will be that the data are in the max-domain of attraction of an extreme value distribution, which makes it necessary to give a brief introduction to the basics of extreme value theory.