Statistical properties of the co-authorship networks from papers from the National Institute of Physics

Abstract

We present a statistical characterization of authors from the co-authorship networks obtained from published papers with the National Institute of Physics as affiliation for the period 2000-June 2016. The distributions are compared with Zipf’s law and with empirical distributions from authors from the American Physical Society (APS), showing the effect of the small network size on the resulting distributions. When co-authorships are multiplied by the number of papers done together, the rank-frequency distributions approach Zipf’s law. On the other hand, unique co-authorships deviate from the Zipf’s law due to the finite number of actual researchers, but show trends that are comparable to the APS data. The results seem to indicate that, even with small and relatively closed groups, the rank-frequency properties observed in larger data sets may still be observed due to the same mechanisms at work in scientific collaboration.