Dear FSL users,
I would like to manually calculate an F-test to be submitted to randomise using the one-sample t-test approach. I am working on COPES values from a first-level model from Feat.
As an example, I basically have two cope images for each subject:
cope10 is a contrast for condition A-C
cope11 is a contrast for condition B-C
There are no groups of participants, so it is a pretty simple within-subjects design (although the actual design used is way more complicated).
What I have done until now is merging all cope10 images for all subjects into one 4d file. This was then submitted to randomise using the one-sample t-test approach. I did the same for cope11. This works fine.
However, as a next step I now would like to calculate an F-test that combines cope10 and cope11, so that I can find brain areas that show a main effect across A, B, and C (i.e. just to look for differences between the 3 conditions).
Is there a way to calculate this F-test manually, perhaps using fslmaths? I forgot to include the F-test in my low-level Feats. Is it possible to submit those values to randmise using the one-sample t-test approach?
What I have tried already is adding cope10 and cope11 in a repeated measures ANOVA model, but I got stuck there with the exchangeability-block information settings because cope10 and cope11 are already basically difference scores (sp randomly inverting the sign is sufficient), but cope10 and cope 11 should be compared within subjects.
Is there a simple way to calculate F-tests or is the only alternative to model the full design with the contrasts for all original EVs?
Thanks for your help.
Best regards,
Henk
Dr. Henk van Steenbergen
Assistant Professor
Institute of Psychology - Leiden University
Leiden Institute for Brain and Cognition (LIBC)
Affect, Motivation & Action (AMA) lab
|