Hi Omar,
> 1. I had originally
> included a third
> group of people who relapsed (so it was a 2x3), but to
> simplify things...I
> will just go for 2x2. Since my all_FA_skeletonised.nii.gz
> already has the
> cases, can I simply put zeros for the cases I want to take
> out and run
> randomize, or do I have to go back and redo the prior steps
> to get a new
> all_FA_skeletonised.nii.gz without the relapsers.
Best is to use fslroi to get an all_FA_skeletonised_control_abstinent with just your first 2x2 groups, no need to re-run anything.
> 2. The design.mat I created has the following:
>
> EV1(control time1)/EV2(control time2)/EV3 (abstinent
> time1)/EV4 (abstinent
> time2)/EV5(subject1)/EV6(sub2)/EV7(sub3)...and so on with
> subjects.
>
> I have tried several simple contrasts, but have not yet
> figured out how to
> do the analysis the way a traditional 2x2 is done with
> simple behavioral
> data. For example, how do I do an omnibus F-test (Omnibus
> ANOVA) to examine
> if there is an interaction effect. What are the actual
> contrasts that need
> to be included, and how do I indicate in Glm that I would
> like an omnibus
> F-test ANOVA.
I think we've already answered to this question last month? If you can't find the answer, please check the archives...
> 3. When an F-test is set up, how are the results displayed.
> Is there an
> output file with specific regions in which these is a
> significant
> interaction effect between the factors?
The F-test and the interaction effect (not answering the same questions) would indeed be displayed as maps with possible significant regions.
I think what you want to do here is an interaction (1 -1 -1 1), not an F-test, but I can be wrong...
> 4. In order to run randomize F-test, is it necessary to
> have equal number of
> subjects in each cell.
Nope
> 5. If there is an interaction, what is the next step, and
> if there is no
> interaction what is the next step?
It depends on your questions.
You might want to do some post-hoc t-tests between your groups (e.g. 1 -1 0 0; 0 0 1 -1) or between your time-points (e.g. 1 0 -1 0; 0 1 0 -1) to understand what's driven your results (or lack of).
Cheers,
Gwenaëlle
|