1

I hope you can help. I am looking for the best way to generate random bipartite graphs with localised structure within one of the node types. Such that type A visits a local group of type B with a particular probability and then within that group connectes an edge with type be also with a particular probablity based on the attribute of nodes of type B. Basically this is social network data where one group has a preference function that determines the probability of linking with another individual of a different type. However the choosey group os restricted in the node value distributions they see due to the localised groupings of the chosen type.

Sorry I hope this question does not bore. All programs I have tried do not allow both this localised structure by trait values as well as probability of pairing based on trait values.

Best Wishes, Colin

Kaveh
  • 21,577
  • 8
  • 82
  • 183
user3776
  • 19
  • 1
  • If you state which programs have you tried, we may be able to point you at others. – Dave Clarke Feb 11 '11 at 14:17
  • Sorry, I have tried Ucinet and Socprog – user3776 Feb 11 '11 at 14:39
  • Really I'm hoping for a very simple agent based model, where i can specify structure in the population in terms of trait value. The results of each time step would need to be realised in netwrok style matrix data. Of who meets who. – user3776 Feb 11 '11 at 15:16
  • Creating random data structures with well-defined distributions is by no means a trivial matter. However, your description seems to imply a very simple model; have you tried coding it yourself? – Raphael Feb 11 '11 at 16:36
  • Perhaps Jung (http://jung.sourceforge.net/) might help. It has various different means for generating random graphs. One of these might suit your needs, or be adaptable. – Dave Clarke Feb 11 '11 at 17:05
  • 1
    Other possibly useful links can be found in this question: http://cstheory.stackexchange.com/questions/2257/network-social-network-analysis-visualization-tools – Dave Clarke Feb 11 '11 at 17:13

0 Answers0