Hi Pengmin,
The mask should not be much bigger than the brain. If
your mask covers the brain fairly well, that is good.
If your mask has 1000000 voxels, and for p=0.05 you have
q=0.9992, that means you have approximately 50010 voxels
above your threshold, say. But the expected number of
false positives is 50000. So the expected fraction of
false positives is 0.9992.
Basically the number of voxels above your threshold is
equal to the expected number of false positives given
the p-value (1000000 * 0.05 = 50000).
You would need to use a higher threshold.
- rick