Hi Nick,

I can't help but think the approach is not a good idea. You list the reasons in your posting and I would add that I don't like that some parts of the resultant beast are 2-manifold and some are not. I am surprised the clustering program does not crash and burn on such objects. You make the point that smoothness estimation does not matter since you're using using bootstrap methods to determine significance, but the ad-hoc method will raise concerns that the results will never hold up to replication.

I would determine with some certainty why these blobs are not reaching significance. Do you have to few subjects? Would other correction methods, such as FDR and FWE fail also? If your maps are quite clean and the blobs fall quite nicely in regions supported by prior independent findings, and you can't get more subjects, you can show the maps without multiple comparison corrections and throw yourself at the mercy of the reviewers.

cheers,
z