I am new to social network analysis, please pardon if this is a simple question.

I am using GEPHI to create a network.

My data is a set of ratings by peers in group settings (sociometric). All peers rated all other peers in the group on a scale from 1-4 that tapped into "liking". Some youth have ADHD, some do not. I would like to use SNA to determine whether kids with ADHD are on the outskirts of the network (less connections) based on less liked.

It seems that the rankings are a weight, but the SNA that I am getting is simply that all kids in a group are connected -- because they all rated each other.

What am I missing here? How can I formulate this model to answer my question better?

Thanks! Andrea


A simple binary network can be constructed as follows: ratings: 1 and 2 treated as 0; ratings: 3 and 4 treated as 1

Now you have a directed binary network, which you can model with an Exponential Random Graph Model (ERGM), in which you can specify many terms of interest, such as

  1. number of edges: this captures the overall density of the network.

  2. number of mutual edges: this captures the tendency of forming mutual edges.

  3. nodematch: this captures tendencies of forming edges within group ADHD and group non-ADHD.

  4. transitivity (GWESP term): captures the tendency of forming edges because they have common connections.

http://www.statnet.org/ provides the statnet R package

  • $\begingroup$ thank you. this is helpful. can you point me in the direction (if it exists) for a statistical test to compare node strength or other statistics across groups - e.g.,to answer if ADHD kids have singificantly less centrality or node stregth than non-adhd? $\endgroup$ Dec 14 '15 at 16:25
  • $\begingroup$ @user3722874 How many kids do you have? I guess not many? I think some t-test can be helpful, for example, compare average rating within the ADHD group and the average rating within the non-ADHD group; compare the average rating of ADHD group towards the non-ADHD group and the average rating of non-ADHD group towards the ADHD group $\endgroup$
    – mac
    Dec 15 '15 at 1:23
  • $\begingroup$ @user3722874 In the directed binary network context, if by centrality you mean "popularity" of each node, then in ERGM, you can use terms: "receiver" and "sender", which captures the number of connections each node received and sent $\endgroup$
    – mac
    Dec 15 '15 at 1:41

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.