help > Semipartial correlation and circularity
Showing 1-4 of 4 posts
Feb 8, 2017 08:02 AM | hannes berg
Semipartial correlation and circularity
Dear Alfonso
thank you for your great toolbox!
I have a question regarding semipartial correlation and circularity. I have entered several networks derived from ICA in seperate seed-to-voxel analyses using bivariate correlation. 2nd level results showed, among others, positive connectivity of those networks to the thalamus. After these results were obtained, analyses were re-run, this time using semipartial correlation and entering the thalamus as a variable to control for.
However, it appeared to me that this analysis is circular and somehow trivial. Is this really the case? I can't still wrap my head around it.
Thank you very much in advance!
Hannes
thank you for your great toolbox!
I have a question regarding semipartial correlation and circularity. I have entered several networks derived from ICA in seperate seed-to-voxel analyses using bivariate correlation. 2nd level results showed, among others, positive connectivity of those networks to the thalamus. After these results were obtained, analyses were re-run, this time using semipartial correlation and entering the thalamus as a variable to control for.
However, it appeared to me that this analysis is circular and somehow trivial. Is this really the case? I can't still wrap my head around it.
Thank you very much in advance!
Hannes
Feb 21, 2017 07:02 AM | hannes berg
RE: Semipartial correlation and circularity
Can anyone give me adive on this?
Thank you in advance
Thank you in advance
Feb 24, 2017 03:02 PM | Shady El Damaty - Georgetown University
RE: Semipartial correlation and circularity
Hi Hannes, I believe that if you are using the seed to voxel
approach on your ICA ROIs then you must selection
multivariate/partial correlation by default in order to properly
replicate what would have obtained in a second level analysis
observed under ICA-->Spatial Summary. I do not think you
need to be worried about circularity because to properly model
brain<->ICA networks you must use the partial/multivariate
correlation.
Mar 10, 2017 08:03 AM | hannes berg
RE: Semipartial correlation and circularity
Dear Shady,
thank you for you reply. I suppose I was not clear enough in my original post. I was wondering if it is possible to use partial correlation in seed-to-voxel analysis to further investigate the findings obtained in this analysis. In my case that would mean to include the thalamus, which was not present as a seed but was obtained as a result of seed-voxel analysis. The idea was to further investigate, which variance in the data is attributable to the thalamic time course. In brief:
1. Seed-to-voxel analysis using ICA networks as seed
2. Result: Several areas show increased connectivity with this seed, the thalamus is among those regions
3. Use partial correlation to regress out the signal of the thalamus (either the whole thalamus or a cluster)
Does that makes sense? Any help would be highly appreciated
thank you for you reply. I suppose I was not clear enough in my original post. I was wondering if it is possible to use partial correlation in seed-to-voxel analysis to further investigate the findings obtained in this analysis. In my case that would mean to include the thalamus, which was not present as a seed but was obtained as a result of seed-voxel analysis. The idea was to further investigate, which variance in the data is attributable to the thalamic time course. In brief:
1. Seed-to-voxel analysis using ICA networks as seed
2. Result: Several areas show increased connectivity with this seed, the thalamus is among those regions
3. Use partial correlation to regress out the signal of the thalamus (either the whole thalamus or a cluster)
Does that makes sense? Any help would be highly appreciated