Hi Anderson,
Sorry for the lack of clarity.
I used randomise to find negative relationships for structure (FA skeleton) and behavior. I did 5000 permutations on a "High" task and a "Low" task separately. My High analysis gave me one cluster that was significant. My low task, however, did not give me any significant clusters. I used the TFCE FWE-corrected image.
My next step was to try to visualize the relationship that randomise found significant in the high task. I overlaid the significant cluster found in the high task onto each subject's FA skeleton and obtained the average FA in that cluster to see if I could find a significant linear regression between the two variables (FA and High task data). I did, which is what was expected.
I then decided to used the same cluster to see if there was a significant linear regression between the "Low" task and the FA values. This also gave me a significant linear regression. What confused me here is that the results are significant when correlating "Low" and FA using R, but no significant clusters were outputted by randomise.
Is it incorrect to take a cluster from one randomise analysis (like the High task), and use it to compare to other behavioral data (Low task)?
Thanks,
Derek Archer
|