Print

Print
Hi Sara,

Even considering that the variance of the interaction is higher, it can indeed happen: a slight increase A to B in one group, a slight decrease A to B in the other, none strong enough, yet their difference is significant, particularly with TFCE, that is more powerful (for combining spatial extent).

But see another comment below:


On 8 May 2017 at 22:48, [log in to unmask] <[log in to unmask]> wrote:
Dear FSL team,

I ran a TBSS to compare scan A and B of two groups (Treatment vs Control). I used the model under "ANOVA: 2-groups, 2-levels per subject (2-way Mixed Effect ANOVA)" from this link:

Then I randomised it using the following command:
randomise -i all_FA_skeletonised -o ANOVA -d design.mat -t design.con -e design.grp -m mean_FA_skeleton_mask -n 5000 --T2

Results show a significant interaction of group*scan while the main effect of scan is not significant.

To dig more into the interaction, I ran four separate TBSS to find the simple main effects. I compared scan A and B in each group separately and then I compared the two groups regarding each scan. But none of the simple main effects came significant.

The second test above (comparison of the two groups regarding each scan) isn't relevant as far as the 2x2 is concerned, as the 2x2 seeks within-subject effects, but the comparison of the groups for the separate scans seeks between-subject effects.

All the best,

Anderson

 
How is it possible?
How can I realize where the significant interaction of group*scan came from?

Thanks,
Sara


On 8 May 2017 at 22:48, [log in to unmask] <[log in to unmask]> wrote:
Dear FSL team,

I ran a TBSS to compare scan A and B of two groups (Treatment vs Control). I used the model under "ANOVA: 2-groups, 2-levels per subject (2-way Mixed Effect ANOVA)" from this link:

Then I randomised it using the following command:
randomise -i all_FA_skeletonised -o ANOVA -d design.mat -t design.con -e design.grp -m mean_FA_skeleton_mask -n 5000 --T2

Results show a significant interaction of group*scan while the main effect of scan is not significant.

To dig more into the interaction, I ran four separate TBSS to find the simple main effects. I compared scan A and B in each group separately and then I compared the two groups regarding each scan. But none of the simple main effects came significant.
How is it possible?
How can I realize where the significant interaction of group*scan came from?

Thanks,
Sara