Peer Influence in the Diffusion of iPhone 3G over a Large Social Network

Our modified version of T-CLAP snowballs from a random adopter with depth d (three, in the case of our paper) and keeps pruning subscribers from this sample while the resulting community has more than s subscribers (115, in the case of our paper). H is the social network graph. H. S is the set of subscribers in this graph and A the set of adopters. The algorithm identifies the subscribers in H that are not adopters with the lowest IER. These subscribers are kept in set V. If V is empty, then the algorithm identifies the subscribers in H that are adopters with the lowest IER. These subscribers are kept in V. This allows us to prune non-adopters with higher probability. A subscriber from V is then selected at random and removed from H. This can, however, separate H into disjoint graphs. If this is the case, then H keeps only the subgraph with the highest IER (ties decided at random).