Hi Michelle,

The [1 1 1 0 0] would be the mean, considering the group differences and cov1 and cov2 as nuisance. It's an unusual behavior that you identify. It doesn't seem impossible but very unlikely, unless some of the significant results you saw earlier were in the negative direction (you may have found that with an F-test), or if they didn't match the same regions in the brain, or if you used spatial statistics (such as cluster extent).

Or there's an error in the interpretation of the other results...

All the best,

Anderson

On Tue, 2 Oct 2018 at 14:21, Michelle Chiu <[log in to unmask]> wrote:

Hello FSL experts,

My level 3 design model was roughly as follows, where copes 1-3 would give me group-level results for my non-overlapping subject groups (GROUP 1-3), respectively (COV1, COV2 stand for covariates).

I later wanted to obtain an overall group activation averaging across GROUP 1,2,3. To do so, I changed my model such that C1 would be (1, 1, 1) for the first three columns. When I ran this model though, I got near zero activation for C1 even though originally, GROUP 1, 2, and 3 all showed significance above .05 threshold for the same contrast.

I was hoping someone could help shed some light as to why this would be the case and how I would set up my model differently to get the group level activation averaging across all three subject populations. Thank you!

Best,
Michelle

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1