Re: [igraph] Welcome to the "igraph-help" mailing list (Digest mode)

[Tamas Nepusz]

Hello,

> Is it possible to rewrite the next two lines
> from
> "if (is.directed(graph)) stop("This method requires an undirected graph")
>     subgraph <- induced.subgraph(graph, vs)"
> to
> "subgraph <- induced.subgraph(graph, vs)
> subg1 <- as.undirected(subg1)"
> ?
Yes, but IMHO it does not make sense - it is effectively the same as
converting your graph to undirected *first* and then testing for the
significance of the communities. By doing so, you are essentially
saying that "the directions of the edges in a community do not
matter". Of course it is fine to say so - but in that case, you could
just run the original community detection algorithm on the undirected
variant of your graph and then test the significance of the
communities with the undirected version of the function (since the
directions do not matter anyway). On the other hand, if you think that
the directions of the edges *do* matter, and you use a community
detection method that takes the directions of the edges into account,
then you *should* test the significance of the communities in the
directed graph as well.

> In the paper
> http://journals.aps.org/pre/abstract/10.1103/PhysRevE.81.046110 or
> https://sites.google.com/site/andrealancichinetti/significant_pre_revised.pdf?attredirects=0
> unfortunately, I couldn't find answer how to apply text to an directed graph
This is because the entire notion of communities in directed graphs is
unclear (at least it was the last time I was following the field
closely). For the undirected case, the authors have argued that
communities in undirected graphs should have more links "within" them
that links leading "out" from them, so a significance test that
compares the within-degrees with the number of links that lead
"outside" makes sense. You need to come up with something similar for
directed graphs if you want to adapt this method.

T.