[jppittma]

I think the fact that nobody knows how to deal with negative correlations in functional connectivity takes away from the validity of the whole concept. People literally throw out negatively correlated voxels/ROI's. This, on top of the fact that functional connectivity is only related to physical connectivity in the abstract. (it's about the correlated time series activity in pixels over time, and isn't bound by any cellular connectivity). I just struggle to discern any meaning from the concept.

[robwwilliams]

I second the question below. The phase, magnitude, and polarity are all used. Yes, sure, you can focus on sub-systems with positive correlation under the presumption that joint increases in signal signify joint increases in O2 use and neuronal activity. Functional connectivity is functional more relevant than mere axonal connectivity. Only a small fraction of an axonal arbor — its collection of presynaptic terminals—-is functionally and reliably coupled to postsynaptic targets.

[free_energy_min]

any source that mentions throwing away negative correlations?

[parpfish]

It’s commonly done in fMRI studies and I always got the impression it was less about not being able to understand negative relationships (e.g., it’s well understood that things like the default mode network is inversely related to active processing) and it’s more about not being clear how to use negative values as edge weights in graph theoretic analysis of networks