Common approaches in connectome studies are to threshold connectivity matrices to produce sparser graphs and to then average edge-level FC estimates between modules/networks, producing estimates of inter-network FC. However, the number of inter-network edges will naturally differ across participants. In the extreme case, one subject may have 1000 edges connecting the Default-Mode and Dorsal Attention Networks that survived thresholding, whereas another subject may have 1 edge that survived. If we take the average score for each of these subjects, I’m sure we can agree that these scores are not equally reliable.
My question is how many edges are needed to provide a reliable measure of average inter-network FC? Does anyone know of any guidelines for this in neuroscience? Or in any other fields - how does the reliability of average scores related to e.g. the number of questionnaire items, the number of trials in a behavioural task or task-fMRI, etc.