Print

Print
Hi Davide,

Please, see below:

On 12 April 2016 at 19:10, Davide1122 <[log in to unmask]> wrote:
Hello everybody,

I'm performing a VBM analysis on three groups.
I look first at simple group differences between the three groups but I don't find results.

This means that the average GM volume is not significantly different between the groups. 
 

I perform then the so called "Two Groups with continuous covariate interaction", adding the performance to tests as covariates as explained in the webpage, and I do find differences between the three groups in function of their performance.

The example in the webpage shows contrasts for the interaction only so I take that the interaction was significant here. This means that the association between GM volume and performance is not the same across the two groups. It can be that in one the association is positive and in the other negative, or both positive but with one significantly stronger, or both negative with one significantly stronger. To find out what is the case, you need to run additional contrasts. Using the example from the page, run:

C3: 0 0 1 0
C4: 0 0 0 1

There is no need to actually do any permutation: only open the test statistic and see the sign for each contrast, and this will tell what is going on.

 

Now, my question is: does it make sense if I comment these results without having any group difference? And how can an interpretation can be?

Yes, it does. The interaction is significant, but not the main effect of group.

All the best,

Anderson

 

Thank you for your reply,

Davide

PS: The two GLM are different and in the first one I don't add any covariate.