I'm relatively new to social network analysis and from what I understand many of the common statistical technique do not work in social network analysis due to the non IID nature of networks. My question is, if I want to determine if there is any link between two values of a node (say gender and the degree or weight of connections and location of the person) how do I do that? Can I use correlation (Pearson, Spearman, etc.)?

  • $\begingroup$ You will want to look up exponential random graph models. For gender for descriptive purposes you could make a simple table, but it would be harder for a continuous variable like location. $\endgroup$ – Andy W Oct 19 '16 at 14:37
  • $\begingroup$ @AndyW Thanks for the response, but aren't ERGM's used to say something like if we are to explain the network with only X property of the network then we can explain Y% of the network structure? Whereas I am not looking to explain network architecture but rather to understand the connection between two properties. I don't know how we would use ERGM's for that. $\endgroup$ – Neal Sidd Oct 23 '16 at 14:11
  • $\begingroup$ No, exponential random graph models predict whether two nodes are connected based on the node properties, so they can assess homophily. It is similar to logistic regression, where the observations are all pairwise combinations of nodes, and those aren't connected have 0's and those that are connected have 1's. $\endgroup$ – Andy W Oct 23 '16 at 14:21
  • $\begingroup$ @AndyW Sounds good, I will look into ERGM's, thank you. $\endgroup$ – Neal Sidd Oct 24 '16 at 3:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.