Robust rankings: Review of multivariate assessments illustrated by the Shanghai rankings

Defined errors are entered into data collections in order to test their influence on the reliability of multivariate rankings. Random numbers and real ranking data serve as data origins. In the course of data collection small random errors often lead to a switch in ranking, which can influence the general ranking picture considerably. For stabilisation an objective weighting method is evaluated. The robustness of these rankings is then compared to the original forms. Robust forms of the published Shanghai top 100 rankings are calculated and compared to each other. As a result, the possibilities and restrictions of this type of weighting become recognisable.