Hello Miguel,
The following is indeed accurate:)
Many Regards
Matthew
>
> Hi FSLers,
>
> I need to explain how randomise + TFCE works in an intuitive manner. Could you please say if the following is accurate?
>
> First of all, an unpermuted t-map is computed. This would be a multiple linear regression at each voxel. TFCE is then fed with this unpermuted t-map, and an unpermuted TFCE-map is computed.
>
> Then, at each permutation in randomise:
> 1. A t-map is computed (and maximum t value is stored for later generating the max(t) null distribution).
> 2. TFCE algorithm is fed with that t-map, TFCE-map is computed (and the maximum TFCE score is stored for later generating the null distribution).
>
> After N permutations, we can compare our unpermuted TFCE-map with the null distribution made of N max(TFCE) scores to get a FWE-corrected p-map.
> (And, if we compared the unpermuted t-map with the max(t) null distribution, we would get voxelwise FWE p-map?)
>
> Thanks a lot,
> Miguel
>
>
