search

chirunconf / chirunconf19

Discussion of potential projects for Chicago R Unconference, March 9-10, 2019
16 stars 2 forks source link

Simulations for Missing Network Data #8

Open benlistyg opened 5 years ago

benlistyg commented 5 years ago

A couple of months ago, I tried finding ready-made ways of simulating missingness in social network data but couldn't find anything that didn't require writing the code from scratch.

I think this would be pretty handy for the network science community (either as a separate package or as a PR / vignette for something like iGraph, statnet, sna).

Kossinets (2008) outlines some potential mechanisms that could lead to missingness in networks that I think would be handy as a starting point to convert to R functions.

alexpghayes commented 5 years ago

Related work (disclaimer: I work with this professor): https://github.com/karlrohe/fastRG. Sampling from sparse matrices with various stochastic blockmodel type setups.

Another related topic we've come across recently is a partially observed network: suppose you know some structural 1s and some structural 0s in a graph adjacency matrix, but part of the network is unobserved. How do you represent this in R? Sparse matrices can conflate structural with missing zeros. How do you load / use these matrices in igraph, tidygraph, etc, etc.

benlistyg commented 5 years ago

@alexpghayes I remember reading (and somewhat comprehending) that paper when Karl tweeted it out! Excited to hear more of your thoughts on this topic this weekend.