I posted this message last week and am still stumped. I appreciate any help you can give.
Dear FSL folks,
I have a question regarding FEAT clustering correction. I used the standard options (z-threshold = 2.3; cluster p-threshold = 0.05) on a third level analysis. This resulted in two clusters of activation: Cluster1 with 695 voxels in the left precuneus/cingulate gyrus, and Cluster2 with 1 voxel in the right cingulate gyrus.
When I view the unthresholded zstat1 image, I see activation spanning across much of the brain, including my two clusters. When I raise the fslview threshold to 4, my two significant clusters disappear, but there is still a lot of activation. If I continue raising the fslview threshold, I still see a lot of activation in the occipital lobe. It seems like this occipital cluster is one with large spatial extent and high z-score.
I don't understand why this occipital cluster isn't surviving the clustering correction. I read that using a low z-threshold gives power to clusters with low z-score and large spatial extent. Using a high z-threshold gives power to high z-score, small clusters. What about clusters with high z-score and large spatial extent? Also, why is my Cluster2 showing up as significant when it's only one voxel in size and has a relatively low z-score (3.38)?
Thanks for the help,