I am doing VBM analysis on SPM. I ran a full-factorial 2 * 2 repeated measure Analysis of Covariance (ANCOVA). I looked at principal effects and interaction for the 2 factors (whole brain for the 3 F).

 

My first question is: I have a-priori regions of interest. For the ANCOVA, is it better to report significant non-corrected whole-brain results (p<0.001), or to apply a-priori masks (small volume correction) and report significant corrected peak voxels? In that case, do I have to correct for the number of applied masks (structures)? And how to correct for that?

 

 

My second question is: What is the best way to do simple effects?

 

I know I can’t consider non-significant structures in the whole brain (or SVC) interaction.

 

Do I have to run t-tests on the whole-brain, and only report structures that were significant in the interaction? Is it better to apply masks on structures that were significant in the interaction?

 

Is it better to do that in SPM, or to extract significant peak voxels and run simple effects with SPSS (for example)?

 

In each of these cases, do I have to apply corrections for the number of t-tests? I don’t think so, because I only consider structures that were significant in the interaction, but I am not sure…

 

For example : In one structure significant in the interaction, my t-tests will be :

-        Factor1 level1 : Factor2 level1 < Factor2 level2; Factor2 level1 > Factor2 level2

-        Factor1 level2 : Factor2 level1 < Factor2 level2; Factor2 level1 > Factor2 level2

-        Factor2 level1 : Factor1 level1 < Factor1 level2; Factor1 level1 > Factor1 level2

-        Factor2 level2 : Factor1 level1 < Factor1 level2; Factor1 level1 > Factor1 level2

 

Thanks a lot in advance,

Sabrina