By examining the literature we found that many of the algorithms used in astronomy as well as in other fields of science are very complicated and need profound mathematical knowledge. In other cases we observed them to be quite simple but not really satisfying in so far as their results are concerned. This experience led to the following three main considerations an algorithm has to fulfill:

- There does not have to be any restrictions concerning the form and dimension of the data set.
- There must be an exact mathematical base of the algorithm.
- The algorithm should be easy to handle in order to enable an application by scientists having a less extensive mathematical knowledge.

The resulting algorithm fully complies with these conditions. The data sets investigated with the algorithm have to fulfill two general conditions:

- There must be a large amount of data forming a total structure.
- There must either be the knowledge that there exists partial structures or the possibility to make an estimation about such partial structures.

Looking at these restrictions the algorithm may be applied in every field of science with large amounts of data. The methodology allows to extract well defined samples from data sets.

*Acknowledgements*

We want to thank R. Weinberger,
J. Pfleiderer and N. Netzer for many valuable discussions.
This work was supported by the Austrian
*Fonds zur Förderung der wissenschaftlichen Forschung*, project number P10036-PHY (DeNIS).

Copyright by the European Southern Observatory (ESO)

web@ed-phys.fr