Re: [igraph] Statistical significance?

[Chris Watson]

I think it will depend mostly on what is common in your field. I think generating a large set of equivalent random networks is generally OK to do. If you are doing group comparisons, permutation/randomization testing would be better, and bootstrapping

On Fri, Jul 3, 2015 at 8:16 PM, Manuel J. Zetina-Rejón <address@hidden> wrote:

Hi everyone

Graph theory is not my expertise area. I’m ecologist working with food webs (networks). I’m usually test hypothesis using non-parametric statistical tests. During the lat year I start using random network models to try to figure out if any result igraph is meaningful, for example, when determining centrality indexes or when identifying communities,  I run Erdos-Reyni with same number of vertices and edges to test if my result is not due to chance. I also use degree.sequence.game( ) similarly.

In your opinion guys, do you think this is enough for testing statistical significance?


Thank you,

Manuel


Ps. Probably this is a widely question or not in the appropriate email list, sorry. But if there any suggestion of additional tests, I expect to run with igraph


_______________________________________________
igraph-help mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/igraph-help