Effect size estimate for tfce corrected randomise results


I was nudged to include effect sizes for my significant clusters yielded by randomise + tfce.

Does it make sense to calculate it using the t-stat outputs (something like 2 x peak coordinate t-stat / sqrt(df))? My hunch is that it does not, since a high t-stat of a given voxel in the uncorrected image may not necessarily yield a low p value in the same voxel in the corrected image.

Also would appreciate alternative ways to show the size of the effect.