Hi Alfonso, thanks again for your reply.

I thought I'd figured it out with your help, but am still a bit confused...

What I'd like to do is find the average connectivity between 14 ROIs, across subjects. As you say below, I can find each subjects correlation matrix in a 14x14x18 (18 subjects) matrix in: /results/firstlevel/ANALYSIS_01/resultsROI_Condition001.mat, so can get average connection strengths by averaging over the 18subs. (This produces the first matrix in the attached image)

I want to then create a graph (I'm using BrainNet for that) only including the connections above an arbitrary threshold, say 0.5, so I just remove all correlations less than 0.5. This leaves approximately 10% of connections. Firstly, does this sound appropriate as I've described?

My problem/confusion is that using the conn gui, when I go into the network explorer window and choose a threshold of 0.5 the ROI graph displayed on the left has considerably more connections that survive this threshold - about 20%. If I export data and load the results.Z (and results.Z_thr) matrices in the .network saved file, average across subs and plot these (also shown in the attached file) the values are different to what I get from the other method. It also doesn't correspond to the graph depicted with some below threshold connections (at least according to the matrix that I have) shown in the image.

I appreciate that to work out the different graph theory metrics graphs need to be computed wihtin each individual and then these statistics averaged over subjects - however, it's not clear to me how to the group graph itself, the one that is displayed, is obtained. For my purposes I'm not actually too concerned with teh Graph theory metrics.

My questions:
1. Does it sound like the method I described first is valid?
2. How is the graph shown in the network theory calculated?
3. What is the cause of the discrepancy between the two - which is 'more' appropriate?

Sorry for the long message!

Thanks!!

Ben