New feature for rdiffnet

[gvegayon]

Allow the model to add more randomness. Instead of setting the adoption rule to be deterministic, i.e. if exposure > threshold then adopt, it should be something like exposure > threshold + eps, where eps ~ F.

[gvegayon]

Also, allow for:

• homophily/heterophily based influence
• preferential attachment-like influence

So the likelihood of adoption it is also based on these.

[hannehuygelier]

Hi,

I would like to apply your R package to predict a sequence of nodes that adopt a certain behavior. So at each time point I need only 1 new adopter. Which adopter that is depends on the diffusion from the node that adopted the behavior at t-1 towards the nodes at t and is probabilistic.

Will you add an option in the adoption rule in which one could sample a single new adopter based on a certain probability distribution?

Kind regards, Hanne

[gvegayon]

That sounds like a nice option to add. In that case difussion would be a rather continuous process as described by a Poisson process. I believe that what you are describing may be available in the RSiena package, as the diffusion dynamics are described as a Poisson process.

I also recommend you to take a look at the EpiModel package from the Statnet group https://github.com/statnet/EpiModel .

I will leave this comment here, but unfortunately, the development of the netdiffuseR package is dormant right now (need more funds to keep going).