Hi,
If you want to run randomise with threshold-based clustering
then the -c ( and -C for cluster-mass ) options are the ones to use.
Many Regards
Matthew
> Dear FSL experts,
> I have been using the randomise program to test for significant
> clusters of
> activation that correlate with a behavioral variable, and have been
> using
> the TFCE option to generate output maps.
>
> If I understand the output correctly, the *tfce_corrp_tstat1.nii.gz
> is an
> image that shows those clusters/voxels that have surpassed the
> corrections
> applied from the permutation results in a 1-p format. I was
> wondering if
> there was any way to determine a cluster-level p-value from these
> output
> maps, similar to what is shown when using the cluster command (with
> the -p
> option). I have assumed that masking the *_tstat1.nii.gz with the
> *tfce_corrp_tstat1.nii.gz image and then running the cluster command
> would
> be incorrect because it relies on assumptions from GRF theory???
> Would the
> proper way to extract a cluster level p be to determine the
> distribution of
> clusters across all permutations, and then determine how many voxels
> are in
> the cluster at the nth percentile? If so, is there a way to output
> this
> information using the randomise command. If not, is there a way to
> compute
> the cluster level p value?
>
> Thanks in advance,
> Eric
>
|