Graphische Darstellung der Anpassungsgüte akteursbasierter Netzwerkentwicklungsmodelle

• Applicant: Professor Dr. Ulrik Brandes
• Subject Area: Theoretical Computer Science
• Term: from 2006 to 2014
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 24614601

A large number of graphical methods have been proposed for the diagnosis of statistical models. The,primary motivation is that such models are often quite restrictive in their assumptions, and it is difficult to assess in which aspects there is an important deviation between an empirical data set and a statistical model supposed to match this data.Graphical methods are present to the human eye features of the data, and features of the datamodel combination, that are not easily discerned by other means and that may point to aspects of the data that are not well represented by the model. The simultaneous graphical exploration of model, data, and fit may lead to finding and fitting a better model.In social network analysis, graphical representations have been of paramount importance from the very beginning, since the complexity of a network structure can often be brought out much more clearly in a visualization than by algebraic properties or numerical descriptive parameters. On the other hand, there are no readily available methods for visualizing evolving networks so as to support reliable model evaluation.The aim of this project is to develop and employ network visualization methods for the interpretation and model diagnosis of statistical models for social network evolution. The models of Snijders [28, and related publications listed in the overall project proposal], which form the basis of our ECRP proposal, are rather algebraic and require complementary graphical methods to be utilized more easily. These methods will be developed in a feedback-loop with both the core modeling project and our substantively and empirically oriented partner projects.

• DFG Programme: Research Grants