Dear Neurostar community,
I have two resting-state connectomes collected under two different rest conditions (from 30 participants). I am interested in quantifying the similarity between these two connectomes.
I’m guessing one way would be to conduct Pearson’s correlation at each node, which would then produce a new matrix displaying the correlation between the two connectomes of each node, but I was wondering if anyone out there as different suggestions as to how one might do this?
If you want to compare the full connectomes I would suggest something like a dot-product of the lower triangle of the two correlation matrices (or even the correlation). You can a shuffle that pairings to generate “random” data to create a null distribution and see if the dot-product/correlation between the two conditions is significant (basically a permutation test).
hope that makes sense
Hi Dr. McIntosh,
thank you very much for your reply. Sorry to follow up, but I am confused as to how to incorporate the connectomes of the 30 participants to get a dot product between the 2 different condition connectomes. Would the procedure you suggested provide the similarity of the 2 connectomes just for a single individual? If I am understanding correctly, for each participant I would vectorize their bottom connectome triangle and compute the dot product between these two vectors (one vector from condition A, the other vector from condition B)?
Thank you for any help with this.
Hi Paul - you are correct that you get one value per subject (product of the two vectorized lower triangles). This would allow you to assess the similarity at the individual and group level thru resampling if you wish.