During the past decades, bibliometrics, and in particular citation analysis, has become an increasingly important tool for research assessment. Citation analysis is regularly carried out at the institutional level, but also at the level of individual researchers, and it often plays a prominent role in tenure and funding decisions. We share the concerns expressed in an important report by the International
Mathematical Union:[R. Adler, J. Ewing and P. Taylor. Citation statistics. Statistical Science, 24(1), pp. 1--14, 2009.] 'The validity of statistics such as the impact factor or h-index is neither well understood nor well studied.' Analyzing the characteristics of citation metrics in a rigorous mathematical framework has become an urgent problem.
In this project, we aim to put citation analyses on a firmer mathematical basis, by proposing stochastic models for citation networks, and by investigating the behavior of citation metrics used today. The stochastics models that we aim to investigate include preferential attachment models and several adaptations of them that include aging and paper potential (that could be linked to the authors of the paper, the journal in which the article has appeared, etc). The metrics that we intend to investigate include PageRank, now predominantly used for the WWW but possibly also useful for citation networks. These metrics step away from measures based on simple counting, such as citation counts and h-index.Further, we aim to qualitatively compare our results on data sets obtained from Google Scholar. For example, in certain domains, we can compare the performance of citation metrics to our own peer assessment. This project also offers several opportunities for `PhD-internships' as there are lots of interesting feats that can be simulated or investigated on the basis of empirical data sets. Our contacts with CWTS (Leiden) may further be helpful to obtain data and give more impact to our results
|Remco van der Hofstad (TU/e) and Gerhard Woeginger (TU/e)
|Eindhoven University of Technology (TU/e)