Dear Andrew,
>
> Although this has been asked several times in the forum, I would like to ask this again to clear up some confusion.
> By using pre-threshold mask option, it is said that 'any multiple-comparison-correction in the thresholding less stringent'
>
> It is up to the user to manually make the threshold less stringent accordingly - is this correct?
> For instance, when using cluster thresholding, the default Z threshold is 2.3 - If I have restricted the search region to ROI mask, would I have to recalculate the z threshold (which will be lower than 2.3 in this case), to make the 'less stringent threshold' work?
typically you still leave the “cluster forming threshold” at 2.3 (or whatever value you usually use). It is when FSL calculates the minimum cluster size that is considered significant that it becomes “less stringent”. Let’s say that for a given “cluster forming threshold” a cluster would have to be 200 voxels big to qualify as signifiant. If you restrict your search to an ROI it might be enough with 100 voxels to be considered significant.
In general I would strongly recommend using randomise (provided you are at the 2nd level) when using small(ish) search region masks.
Jesper
>
> If not, I am confused as to where the less stringent thresholding is happening.
>
> Thanks in advance,
|