Dear experts,
I am trying to use randomise to identify clusters of voxels that survive cluster correction. I have tried several of the options (-C, -c, -x, and -T). My question is:
Say I use a threshold (e.g. -C) of say 3.1 or 2.6 when calling randomise. E.g.: randomise -i $input -o $output -d design.mat -t design.con -m $mask -C 2.6.
Can I then, when *viewing the output file* (*clusterm-corrp*.nii.gz), set the minimum threshold to 0.95, i.e. corresponding to an alpha of 0.05, or do I have to also set it to the corresponding alpha for those thresholds, i.e. 0.999 for z = 3.1 or 0.99 for z = 2.6? Is there a way to threshold the activations e.g. at z = 2.6, but to look at a probability of only 0.05 of obtaining a cluster thresholded at 2.6? In other words, if I threshold the activations at one level, can the minimum threshold setting when viewing the output (corrp) file be set at a different level?
I am trying to obtain e.g. a cluster where each voxel survives z = 2.6, but the cluster itself survives 1.7. I'm wondering if I'm misinterpreting something in the instructions for randomise.
Many thanks,
Yoss
|