Print

Print


Hi Jie Yu,

Thanks for sending. Design and contrasts look fine. A significant C3 means that the slope for group 1 (along time) is higher than for group 2, whereas C4 means the opposite.

A significant C3 can happen if:
- There's an increase along time for group 1, and a decrease for group 2.
- There's an increase along time for group 2, and an increase even higher for group 1.
- There's a decrease along time for group 1, and a decrease even more pronounced for group 2.

To disambiguate these, you'd need to investigate the signs for each group (no need to test significance, just check the signs). For this, it's simpler to use a slightly modified design -- please see attached. Here EV1 codes the ages for group 1 only, EV2 for group 2 only. With this design:
- C1 and C2 test for changes along time for both groups (equivalent to your current C1 and C2).
- C3 and C4 test for the interactions (equivalent to your current C3 and C4).
- C5 and C6 are not meant to be used to test significance, but just to check if the voxels have positive or negative signs.

Hope this helps!

All the best,

Anderson


On 28 November 2014 at 15:56, Jie Yu <[log in to unmask]> wrote:
Dear Anderson

Please see the attachments.

Just a reminder, I am basically copying this design:
http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-groups.2C_2-levels_per_subject_.282-way_Mixed_Effect_ANOVA.29


Thanks!!!