Hi Mark,
Thanks for your feedback. I was hoping to illustrate the nature of the correlation, as identified in those voxels that were included in the GLM. If I extract the mean in the way which I described earlier, then plot the mean GM in that cluster against the behavioural covariate, I’m not convinced that the mean is capturing the relationship very well. If I take the first eigenvector using fslmeants, the data look more as I would have expected (this expectation being on the basis of the outcome from randomise).
Can you see any issues then in using the principle eigenvector?
Sam
|