Agarwal, Deepak (2003), Bayesian Models for Sparse Edge Weighted Directed Graphs, Computing Science and Statistics, 35, I2003Proceedings/AgarwalDeepak/AgarwalDeepak.presentation.pdf ,
I2003Proceedings/AgarwalDeepak/AgarwalDeepak.presentation.ppt

We propose a new class of models based on Stochastic Blockmodels that provide global measures for a directed graph based on local interactions. The models we implement differ from the ones that already exist in the literature that focus on very small (20-30 nodes) unweighted graphs that are not too sparse. Our models apply to large (200-300 nodes), extremely sparse weighted graphs. The issue of sparseness is tackled by building Bayesian models that are known to be computationally intensive. The models are fitted using an E-M algorithm which has performed well so far. We illustrate our methodology by fitting the models to some subgraphs of a large telecommunications network.