Hi Greg,

There may be a more efficient answer, but one solution I can think of would be to run a 2nd level fixed effects model on Group B subjects. You can put all subjects into a single model and your design matrix will simply be an identity matrix and you can create a cope for each subject. The cope and varcopes from this fixed effects model should be equivalent to the copes and varcopes from level 1, but in standard space (in an analysis I just ran they were the same). Then you'll have something that the level 3 group model should be happy with.

If you want to verify that what I'm saying is correct, you can compare the 2nd level filtered_func data (4D image of standard space first level copes) and var_filtered_func_data (similar but lev 1 varcopes) to the copes and varcopes from the lev 2 analysis.

Cheers,
Jeanette

On Sun, Aug 24, 2008 at 12:10 PM, Greg Burgess <[log in to unmask]> wrote:

Hi FSL list,

I have two groups of subjects that were each run in a similar task but with slightly
different task designs. Group A had three separate MR runs with 108 trials per run, but
Group B had one run with 324 trials. I wanted to compare my lower-level contrasts
across the two groups in an unpaired t-test, but the higher-level Feat failed twice, in
different ways. If someone can tell me how to compare these two studies, I'd be
appreciative.

For Group A, I ran one lower-level analysis for each run, and a second-level fixed effects
analysis to combine the three separate runs. For Group B, there was only one run, so I
only ran a lower-level analysis. I then created a higher-level analysis and input the 3D
cope images for each subject from GroupA.second.level.feat/stats/copeA.nii.gz and GroupB.first.level.feat/stats/copeB.nii.gz (the cope numbers for identical contrasts were
different in the two groups).

It appears that the higher-level feat will not run because the cope images are in different
image spaces. Group A's copes are in standard space because of the second-level
analysis that combined the individual runs. Group B's copes are in BOLD space because
they are first-level analyses. However, even if I use GroupB.first.level.feat/reg_standard/stats/copeB.nii.gz as the input, the feat fails because
it can't find GroupB.first.level.feat/reg_standard/design.fsf

Can someone offer a suggestion?
Thanks,
--Greg

____________________________________________________________________
Greg Burgess, Ph.D.
Research Associate, Institute of Cognitive Science
University of Colorado - Boulder
Phone: 303-735-5802
Email: [log in to unmask]

University of Colorado - Boulder
UCB 594
Boulder, CO 80309-0594