help
help > RE: Voxel degree centrality maps
Jun 15, 2016 06:06 PM | Alfonso Nieto-Castanon - Boston University
RE: Voxel degree centrality maps
Hi Jeff,
I am not entirely sure if you are referring to the sign of the correlation values. If so, then ICC is sign-independent (since it considers r-squared values), while global correlation is not (it averages the actual r correlation-coefficient values, considering their sign, so positive and negative correlations will cancel out). For ROI-to-ROI graph analyses you can actually choose whether you prefer to disregard the signs (selecting "two-sided" when defining the adjacency matrix will keep the highest absolute-correlation connections, i.e. treating r=0.5 and r=-0.5 equally) or not (selecting "one-sided" there will keep the highest positive-correlation connections only). If, on the other hand, you are referring to whether those r2 and r values represent actual Pearson correlation coefficients (e.g. not Fisher-transformed correlations) then yes, they are the actual correlation coefficients (or their squared values) which get aggregated across voxels in the case of ICC and Global Correlation degree centrality measures. For ROI-to-ROI analyses again you can choose whether you want to threshold based on actual Pearson correlation values, Fisher-transformed values, or cost measures when defining the graph adjacency matrices.
Hope this helps
Alfonso
Originally posted by Jeff Browndyke:
I am not entirely sure if you are referring to the sign of the correlation values. If so, then ICC is sign-independent (since it considers r-squared values), while global correlation is not (it averages the actual r correlation-coefficient values, considering their sign, so positive and negative correlations will cancel out). For ROI-to-ROI graph analyses you can actually choose whether you prefer to disregard the signs (selecting "two-sided" when defining the adjacency matrix will keep the highest absolute-correlation connections, i.e. treating r=0.5 and r=-0.5 equally) or not (selecting "one-sided" there will keep the highest positive-correlation connections only). If, on the other hand, you are referring to whether those r2 and r values represent actual Pearson correlation coefficients (e.g. not Fisher-transformed correlations) then yes, they are the actual correlation coefficients (or their squared values) which get aggregated across voxels in the case of ICC and Global Correlation degree centrality measures. For ROI-to-ROI analyses again you can choose whether you want to threshold based on actual Pearson correlation values, Fisher-transformed values, or cost measures when defining the graph adjacency matrices.
Hope this helps
Alfonso
Originally posted by Jeff Browndyke:
And, those r2 and
r-correlation values are absolute correlation values, right
Alfonso?
Warm regards,
Jeff
Originally posted by Alfonso Nieto-Castanon:
Warm regards,
Jeff
Originally posted by Alfonso Nieto-Castanon:
Hi
Vinay,
There are a few options in CONN for analyzing degree centrality of either ROI-to-ROI graphs or voxel-to-voxel graphs. First, for ROI-to-ROI graphs, you may simply click on the 'graph theory' button on the ROI-to-ROI second-level results tab, enter there the network-forming threshold of choice for defining your graphs' adjacency matrix, and then select the "degree" or "cost" measures to evaluate the degree centrality of the network or of each ROI within those networks. For voxel-to-voxel graphs, in the voxel-to-voxel first-level analysis tab you may select there either the "Intrinsic Connectivity" measure or the "Global correlation" measures. These compute slightly different degree centrality measures for each voxel. The former computes for each voxel the correlation between this voxel timeseries and every other voxel in the brain and then computes the average of those squared-r correlation values, while the latter also computes for each voxel the correlation between this voxel timeseries and every other voxel in the brain but then computes instead the average of those actual r correlation values (not squared).
Hope this helps
Alfonso
Originally posted by vinay gupta:
There are a few options in CONN for analyzing degree centrality of either ROI-to-ROI graphs or voxel-to-voxel graphs. First, for ROI-to-ROI graphs, you may simply click on the 'graph theory' button on the ROI-to-ROI second-level results tab, enter there the network-forming threshold of choice for defining your graphs' adjacency matrix, and then select the "degree" or "cost" measures to evaluate the degree centrality of the network or of each ROI within those networks. For voxel-to-voxel graphs, in the voxel-to-voxel first-level analysis tab you may select there either the "Intrinsic Connectivity" measure or the "Global correlation" measures. These compute slightly different degree centrality measures for each voxel. The former computes for each voxel the correlation between this voxel timeseries and every other voxel in the brain and then computes the average of those squared-r correlation values, while the latter also computes for each voxel the correlation between this voxel timeseries and every other voxel in the brain but then computes instead the average of those actual r correlation values (not squared).
Hope this helps
Alfonso
Originally posted by vinay gupta:
Hello Arkan !
I am also trying to plot voxel degree centrality graphs using CONN. I am wondering whether CONN really has some feature which enables us to plot a degree centrality graph directly ?
I have spent a lot of time researching on this but still have not stumbled upon any reliable source to know CONN specs and the fact that whether it could support this or not.
It would be great if you could share your experience and knowledge, like how you overcame these difficulties ?
Regards
Vinay
I am also trying to plot voxel degree centrality graphs using CONN. I am wondering whether CONN really has some feature which enables us to plot a degree centrality graph directly ?
I have spent a lot of time researching on this but still have not stumbled upon any reliable source to know CONN specs and the fact that whether it could support this or not.
It would be great if you could share your experience and knowledge, like how you overcame these difficulties ?
Regards
Vinay
Threaded View
| Title | Author | Date |
|---|---|---|
| Arkan A | Sep 29, 2015 | |
| vinay gupta | Jun 14, 2016 | |
| Arkan A | Jun 16, 2016 | |
| Alfonso Nieto-Castanon | Jun 15, 2016 | |
| Jeff Browndyke | Jun 15, 2016 | |
| Alfonso Nieto-Castanon | Jun 15, 2016 | |
| vinay gupta | Jun 16, 2016 | |
| Jeff Browndyke | Jun 15, 2016 | |