CollEc documentation

Background

The RePEc Author Service (RAS) allows for authors to identify, or 'claim', papers they have written in the RePEc digital library. When two authors claim the same paper, they are co-authors. The relationship of co-authorship creates a network between most authors. The shortest paths in that network between different pairs of authors are of different lengths. Between co-authors, the shortest path has length one; when two authors have each written with a third author, but have not written a paper together, the shortest path relating those two is of length two, and so on.
There must be at least one author for whom the average length path to all others in the network is the minimum. This author will have the highest closeness ranking.
There must be at least one author who appears on the largest number of shortest paths between any two other authors. This author will have the highest betweenness ranking. A number authors are marginal, i.e., they are not on a shortest path between any other pair of authors. CollEc does not report a betweenness rank for them.

Network type

This is a binary network. Two authors are linked if they have claimed at least one paper in common in RAS. CollEc calculates paths using an extension of Dijkstra’s algorithm. The extension allows for the calculation of all shortest paths between two authors, rather than just one as the standard algorithm would do. For some couples of authors, this leads to large sets of shortest paths. Shortest path multiplicity would clog the user interface of CollEc. Fortunately most of the multiple binary shortest paths can be eliminated when we take collaboration strength as a secondary criterion. The elimination of these shortest paths does not impact closeness centrality rankings but it does impact betweenness centrality.

Status

This site is maintained by me, Thomas Krichel. It should be fully functional and usable. If you notice a bug or have questions, please get in touch. Note that the site will never be fully accurate, because the calculations involved take a very long time. The results will therefore always lag over the state of the co-authorship network.