Dear Prof Zalesky,

I have noticed that when analysing relatively large networks (e.g >300 nodes) the time taken for analysis does not increase linearly e.g. time for 1000 perms>> time for 10*100 perms. I therefore planned to parallelise in a very basic manner by just e.g. running 50*100 permutations rather than a one off 5000 permutation run, and averaging the pvalues.

However it looks as though there may be a problem with this as when I run 100 permutations several times the pvalues are identical, suggesting to me that it is the same permutations occurring each time rather than being chosen at random. This also occurs even if I try setting different random seeds.

Do you know what might be going on here, both why there is the nonlinear increase in time taken to run permutations, and why the permutations appear to be identical when ran several times?

Thank you!

Rob