On Wed, Apr 13, 2011 at 11:00 PM, Matthew P. White <[log in to unmask]> wrote:

Hello FSL Gurus,

I was interested in exploring randomise (specifically TFCE) on some lower level Feat analysis for event-related fMRI data. More specifically, I have two runs of a similar task for each subject on which I have done a second-level fixed-effects merge (mean) for each subject, using Feat.

1) Is it possible to feed cope images into randomize if they are the result of a second-level FE mean? If so, what file would I feed: the cope file?

They must be for a second-level (i.e., group-level analysis), as randomise-TFCE does need a design matrix, whose exchangeability provides a tool for generating null TFCE distribution ...

If you do so, you concatenate a series of COPE images and make the 4D input to randomise (i.e., 4th dimension being subject ID, and remaining 3 being 3D COPE)

2) Or do I need to feed in the lower level Feat directories? Same question as to files.

3) If #2, is it possible to merge two runs (not concat) before doing higher level analysis in randomise?

Thank you for our answer and maintaining this forum.

Matt

Matthew P. White, M.D.
Instructor, Psychiatry
401 Quarry Road, Stanford, CA 94305
650 725-5598