I have a network where nodes are labelled in two classes "a" and "b". I want to measure how connected these two groups are and I looked at assortativity by group. I want to use this measure in more than one network and I d like to compare this metric between different networks (in any network nodes are labelled a or b). my idea is that for low assortativity it means that the two groups are connected enough between each other
however, I m not sure if this is the best thing to do as different networks might have different structural properties.
In particular there are two factors that are worrying me: how interconnected group a is with itself and how interconnected group b is with itself.
given the same number of edges between people in group a and group b assortativity increases if I reduce the number of edges within group a and within group b. this makes me think that even if I get low assortativity I cannot infer that group a and group b are well connected as it could mean that they are generally poorly connected.
alternatively, I was thinking to use Louvain algorithm to cluster the network and use Normalized Mutual Information given the louvain clusters and groups (a and b).
any thought on this?