Re: [igraph] estimating parameters for use in a barasi game

[Tamas Nepusz]

Hi Jesse,

> I have some empirical networks that I need to use as a basis for a
> monte-carlo simulation.  I need to generate new networks that have
> similar characteristics to the current networks (i.e. matched on
> density, in and out degree distributions).  The empirical networks
> exhibit preferential attachment, so I am using the "barabasi.game"
> functions to generate the networks in R, but I am having a hard time
> getting the input parameters right so the degree distributions of the
> empirical and generated networks are consistent
I think it's really hard to achieve that even if your original network
exhibits preferential attachment. I would try a different approach, drop
the assumption of the network being a Barabasi-Albert network and simply
generate my simulated networks using degree.sequence.game (in R) or
Graph.Degree_Sequence (in Python) with the Viger-Latapy sampling method.
This method is said to sample uniformly from the space of networks
having the same in- and out-degree distribution. Another way to go is to
take your original network and rewire it while preserving the degree
distribution; this is implemented in rewire (in R) or Graph.rewire (in
Python). Of course you will have to use a large number of rewiring
iterations to make your resulting simulated graph more or less
independent from your original.

-- 
Tamas