Dear Matthieu,
there is no difference in the handling of height and cluster (extent) thresholds between voxel- and surface based data. You can either correct for multiple comparisons using height or extent (or even both, which is not really necessary). The most often used approach is to report and to correct for height, but of course it makes sense to additionally or exclusively apply an extent threshold to rather focus on effects with larger extent. There were many discussions about that topic with a peak in the last months due to a PNAS paper which raised some doubts about the validity of cluster thresholds (short answer on this topic is that you should prevent using an initial height threshold that is more liberal than p<0.001). Search for "cluster failure article" in the spm maillist and also check previous discussions about the use of cluster thresholds.
Best,
Christian
On Wed, 9 Nov 2016 22:04:07 +0100, Matthieu Vanhoutte <[log in to unmask]> wrote:
>Dear CAT12’s experts,
>
>Could you explain some of the basics of thresholding statistical maps at surface level with the « Transform SPM maps - surface » tool ?
>
>All maps can be thresholded using height and extent thresholds: does it mean that the most reliable result come from both FWE corrected (p<0.05) height and extent thresholding ?
>
>What is the relative value of uncorrected (p<0.001) peak level threshold and FWE corrected (p<0.05) extent threshold ?
>
>Tanks in advance for clarifying it to me !
>
>Best regards,
>Matthieu
|