Hello,
Design 1 is the correct one here - the group column is not relevant for randomise with this design. Your first set of contrasts is probably not what you want - they will just find voxels where any group has non-zero FA. The second set of contrasts does look for group differences - you only need an F over two of the contrasts (e.g G1 > G2, G2 > G3 ) to check if there are _any_ significant differences - you can then establish directionality for significant voxels with the individual contrasts.
Hope this helps,
Kind Regards
Matthew
> Dear FSL-experts,
>
> I am new to FSL and tried reading up on this issue on the FSL mailing list and the tutorials, but got a little confused as to what is the right method for my purpose. I would like to compare 3 groups of patients (no healthy control group as reference in my design) on their FA values in TBSS. Assuming I have 2 participants in each group, my model looks like this:
>
> Group Group 1 Group 2 Group 3
> 1 1 0 0
> 1 1 0 0
> 1 0 1 0
> 1 0 1 0
> 1 0 0 1
> 1 0 0 1
>
> If I understood the FEAT tutorial correctly, the group column should be all 1’s or is that incorrect and it should rather be:
>
> Group Group 1 Group 2 Group 3
> 1 1 0 0
> 1 1 0 0
> 2 0 1 0
> 2 0 1 0
> 3 0 0 1
> 3 0 0 1
>
> I found both information on the tutorial and am confused as to which one is the right one.
>
> http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FEAT/UserGuide#F-Tests OR http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#F-Tests_.28Inter-Group_differences.2C_no_repeated_measures.29
>
> I would like to run an F-test first and follow up a significant result using pairwise t-tests. Following the tutorial, I set up the following contrasts for the F-test:
>
> contrasts F-test
> Group 1 1 0 0 x
> Group 2 0 1 0 x
> Group 3 0 0 1 x
>
> I would calculate this first and then set up a second analysis with t-contrasts after receiving significant results. Or would it be better to set up the contrasts as follows:
>
> contrasts F-test
> X Gr 1 > gr2 1 -1 0 x
> X Gr1>gr3 1 0 -1 x
> X Gr 2 > gr3 0 1 -1
> X Gr 3>gr2 0 -1 1
> x Gr 2 > gr1 -1 1 0
> X Gr 3>gr 1 -1 0 1
>
> However, in the latter case I am not sure which contrasts to choose as F-tests, since I do not have any control group for which I would assume higher FA values.
>
> Any help would be highly appreciated! Thank you!
>
> Bianca Kollmann, PhD candidate
> Department of Clinical Psychology and Neuropsychology
> Johannes Gutenberg-University Mainz, Germany
|