Dear Alfonso and forum members,

I am currently working on a voxel-to-voxel connectivity analysis to assess potential correlations between connectivity values and a symptom scale.

I decided to compare the intrinsic connectivity and global correlation methods. According to the CONN manual, they are both measures of network centrality at each voxel and the former output is the root mean square of the correlation coefficient values while the latter provides the average of the correlation coefficient values. Therefore, I was not expecting my results to vary widely depending on the approach. However, they are completely different regardless of the contrast explored.

For example, a specific result of the contrast AllSubjects(0), SymptomScale(-1), Age(0), Sex(0) that survived the FWE correction using the ICC approach and had a cluster size of 300 voxels does not even appear when using the Global Correlation, at a significance level of p<0.001 uncorrected.

Can someone help me elucidate the reasons why my results are so different when using ICC or Global Correlation?

Thanks kindly, and I am sorry if this is mentioned somewhere obvious.

Best regards,
Maria