Dear Eugene,
Thanks for the reply. However, I am not sure I am following.
< Setting a higher z-threshold makes it less likely that clusters will survive this thresholding by chance.>
Yes, and that's exactly why I am scratching my head about this. How come a cluster might (occasionally) become significant when I set a higher cluster-forming threhold (e.g., z=2.58), and not if I set a lower cluster-forming threshold (e.g., z=2.3), while keeping the cluster level p thresholds identical (e.g., p=0.05 in both cases)? Shouldn't a higher threshold make a cluster smaller, thus increasing its p-value and making it less likely to survive cluster size thresholding?
< Thus, those that do will be assigned a smaller p-value, and there will be a smaller cluster-size threshold.>
What do you mean by "and there will be a smaller cluster-size threshold"? Do you mean that the cluster-forming threshold affects the cluster-size threshold, such that the minimum cluster size considered significant is larger when smaller cluster-forming thresholds are selected (e.g., 200 voxels if I choose z=2.58; 210 voxels if I choose z=2.3) even if I keep the cluster p threshold identical?
Thanks,
Marco
|