 |
 |
Dear Anderson,
Awesome, thanks a lot! However, I may need a little clarification since I could not find a similar design in the FSL mailing list:
Subject and EB are clear. See should be run or intervention in my case, and EV1 and EV2 belong to the first group, but what are they modelling?
The rest is pretty clear.
Thanks a lot
Yacila
Von: FSL - FMRIB's Software Library [mailto:[log in to unmask]] Im Auftrag von Anderson M. Winkler
Gesendet: Freitag, 29. Juli 2016 09:33
An: [log in to unmask]
Betreff: Re: [FSL] Repeated measures for IC in randomise or PALM
Hi Yacila,
Please, see below:
On 28 July 2016 at 16:06, Yacila Isabela Deza Araujo <[log in to unmask]> wrote:
Dear FSL experts,
Dear Anderson,
Now I am trying to run the last stage of my analysis with 2 groups and 3 interventions.
So far, I think I will follow this design, plus one more run: http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-groups.2C_2-levels_per_subject_.282-way_Mixed_Effect_ANOVA.29
Because it is 2 groups (between subjects), 3 levels (within subjects) design.
Please, see a pair of designs with all possible contrasts at this link:
My specific question is: How can I proceed in PALM. I think I may have thinks clear for randomise, but should be PALM different?
For the within-subject effects or between-subject effects (separately), randomise can be used (within-block permutation for the within; whole-block permutation for the between). PALM will add the possibility of running both simultaneously, and correct across all designs and contrasts.
My other question is related to the amount of data: I want to test IC from MELODIC which should be entered as different modalities (I think), and I would like to use the fast inference method, since I have a total of 80 subjects * runs = 240 IC.
So 240 volumes, times a number of ICs (modalities). The call would be something as this:
palm -i dr_stage2_ic1.nii -i dr_stage2_ic2.nii -i dr_stage2_ic3.nii [...] -d design_within.mat -d design_between.mat -t design_within.con -t design_between.con -eb EB.grp -within -whole -corrcon -corrmod -nouncorrected -n 500 -approx tail -T -logp -o myresults
Additionally, I would like to know if I can model (post-hoc) linear or quadratic contrasts because my interventions are: antagonist-placebo-agonist and in case I have main effects, I would like to know the direction.
I'm not sure I follow. In any case, the contrasts as in the spreadsheet above are directional (all of them), and the results will be corrected across all (with -corrcon). These are already what would be called post hoc, and one can look directly into them.
The F-tests indicated in the spreadsheet aren't being executed in the command line above, as these aren't really useful since you will look into the (already corrected) t contrasts.
Hope this helps.
All the best,
Anderson
Thanks again for your kind help,
Yacila
--
Yacila Deza Araujo, M.Sc. Neuropsych.
PhD Student
Technische Universität Dresden
Faculty of Medicine Carl Gustav Carus
Department of Psychiatry & Psychotherapy
Section of Systems Neuroscience
Würzburger Straße 35
01187 Dresden
Germany
Phone: +49 (0) 351 463 - 42300Fax: +49 (0) 351 463 - 42202
mailto: [log in to unmask]