Tamas, thanks a lot for your advices.
> Yes, I strongly suspect so. I do agree. > Well, as a first approximation, you can say that nodes with a low coreness > value are the periphery and the rest is the core. Alternatively, you could > say that the big strongly connected component is the core and the rest is the > periphery. You could also try to fit a stochastic blockmodel to the network > with two groups -- this is not implemented in igraph, but I have a working > implementation for both traditional and degree-corrected stochastic > blockmodel fitting in C++ (using igraph) so I can help you with that. The > source code is here in case you are interested: Agree, but the problem is how to select the core threshold that determines to be in or out the core... anyway, here's how I proceeded: I made the network symmetric using as.undirected(), using "mean" to average the weights. this way my links mean reciprocal cooperation. This leave the network with a lot of isolated nodes and one big component. I have thus deleted all the isolated nodes, and focused on the big connected component. Now edge.betweenness.community() reveals some meaningful communities. Do you think such an approach is sound? Of course, your suggestion looks really interesting and I will give it a try, but I have no knowledge of C++, so I don't know how to manage such an analysis... should I use the C++ igraph? or can I call the C++ code from R in a relatively easy way? best, Simone -- Dr. Simone Gabbriellini DigitalBrains srl Amministratore Head R&D mobile: +39 3403975626 mail: [email protected] _______________________________________________ igraph-help mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/igraph-help
