Dear CONN community,
I'm conducting ROI-to-ROI FNC analyses for evaluating between
group differences. I'm adopting a hierarchical clustering procedure
to identify networks of ROIs with a data-driven approach. I've
understood that clusters are defined based on ROI-to-ROI anatomical
proximity and functional similarity. However, I was wondering how
the tool determines the number of clusters being defined and used
in downstream analyses.
I've looked into the manual and the forum but I didn't find any
information. Sorry if the topic has been already discussed.
Any insights would be greatly appreciated.
Thanks,
Ludovico
Hi Ludovico,
The number of clusters is chosen using a classical "elbow" method (plotting the largest within-cluster distance as a function of the number of clusters and selecting the point within minimum x+y normalized coordinates in that plot). You can manually change the number of clusters by selecting the button "ROIs sorted using hierarchical clustering" in the GUI and then selecting the option "use hierarchical clustering method (default)" and entering there the desired number of clusters (or leaving that empty to let CONN select that number using the elbow method above)
Hope this helps
Alfonso
Originally posted by Ludovico Coletta:
Dear CONN community,
I'm conducting ROI-to-ROI FNC analyses for evaluating between group differences. I'm adopting a hierarchical clustering procedure to identify networks of ROIs with a data-driven approach. I've understood that clusters are defined based on ROI-to-ROI anatomical proximity and functional similarity. However, I was wondering how the tool determines the number of clusters being defined and used in downstream analyses.
I've looked into the manual and the forum but I didn't find any information. Sorry if the topic has been already discussed.
Any insights would be greatly appreciated.
Thanks,
Ludovico