Dear Dr. Alfonso, thank you for your help in the forum. I am wondering what kind of inference can I make if I had no a priori hypothesis and just selected all the ROIS in the brain to a Roi-to-ROi analysis in a study where I am comparing pre and post intervention of a Treatment vs sham. I chose the cluster base inference - Functional network connectivity. I attached the images from my analysis for clarity purposes. I am not sure what can I conclude from this image:

Can I conclude that between each Roi I see, there is a greater connectivity ( in the case of red lines - since the contrast I used was -1 pre 1 post and 1 treatment -1 sham) change in the treatment group? or since it is a cluster inference I can only conclude that there is somewhere in the cluster a greater connectivity but cannot conclude this for each ROi individualy? In other words can I say these areas are more connected after the treatment or am I only aloud to say that somewhere in these areas there is a true greater connectivity?


Thank you for your Help, 


Marco Echevarria

Dear Marco,

Yes, since these are cluster-level statistics you typically will want to phrase the results at the same level (i.e. groups of connections, rather than individual connections). For example, if I am understanding correctly, your analyses are looking at a treatment-by-type interaction (e.g. are the changes in connectivity between pre- and post- treatment different when comparing the treatment vs. the sham groups?), and you are finding three significant clusters of connections that show such interaction. These connections could be described, for example, as: a) connectivity between DMN (Posterior cingulate / Precuneus) and SensoriMotor areas (pre- and post- central gyrus); b) connectivity between DMN (Posterior cingulate / Precuneus) and Auditory sensory areas (Heschel's Gyrus, Planum Temporale, Planum Polare, etc.); and c) connectivity between medial (Occipital Pole) and lateral (Occipital and Temporo-Occipital Fusiform Cortex) visual cortices (keep in mind that any similar-level summary of the connectivities involved would work just as well, not necessarily constrained to describing these as three separate groups -e.g. connectivity between precuneus and central sulcus areas including auditory, motor, and somatonsensory areas might work perfectly fine to describe two of those clusters).

Then, regarding the interpretation of the effects (how those connections are being modulated by treatment), the red/blue colors only indicate the general sign of the change (e.g. is the change in connectivity pre- vs. post- treatment greater/more-positive or smaller/more-negative in the Treatment compared to the Sham group?), but that is not terribly informative on itself. I typically find that it is much better to use the "display effects" button (the one with the graph/errorbars image) in order to help you interpret the results in terms of the original connectivity strength values. For example, in this analysis, the 'display effects' button will display the average connectivity values within each of these three clusters for each of the four relevant conditions (e.g. average connectivity in pre-treatment condition within the Sham-group, post-Sham, pre-Treatment, and post-Treatment). You can use those plots to better interpret what the interaction effect means within each of these three clusters

Hope this helps
Alfonso
Originally posted by Marco Echevarria: