On 15 Feb 2012, at 20:40, Anna wrote:
Hi there,
I am running a patient vs. control comparison. FEAT with cluster thresholding was missing small subcortical effects I was looking for, but Randomise identified significant clusters in the ROIs I hypothesized. To create a t-map for these clusters, is it appropriate to simply select voxels above an a prior t-threshold (e.g. 2.3) that fall within the clusters that Randomise identified?
I'm not quite sure what you're asking for. If you have applied an a priori mask for the randomise analysis, then anything in the *corrp* image that is significant can be reported on, including using that as a mask to apply to (eg) the raw tstat image, if you wanted to report the tstats at the voxels surviving the randomise thresholding.
I would also like to look at the correlation of activation in these voxels with clinical variables. Whilst there is a tutorial to do this in FEAT, is there a way of using the voxels I identified from the Randomise analyses I described above?
You can use the mask of significant voxels against the data that was input to randomise, to get data to correlate against clinical variables.
Cheers.
Thanks in advance for your thoughts
Anna