Reply-To: | | [log in to unmask][log in to unmask]> wrote:
Dear Peter,
Thank you for your reply. So just to confirm based on what you said:
1) Is it fine if I use specific contrasts in my GLM to extract the VOIs (e.g. cond1 vs cond2 only), but then I specify different contrasts in DCM and PEB (provided that the 6 conditions are always the same)?
2) Is it best to model all 6 conditions in my DCM instead of leaving 2 conditions out as implicit baseline? I could include all 6 conditions as driving input (the "task" condition which will be cond1+cond2+...+cond6) then only include 4 of those conditions as modulatory inputs in B. Does that sound better? I will then build my PEB contrasts on those 4 conditions in the B matrix.
Best wishes, Sara
On 25/07/2022, 11:14, "Zeidman, Peter" <[log in to unmask]> wrote:
Dear Sara Sorry for the delay in replying.
> My first question is: is it ok to extract different VOIs from different contrasts (granted that they are adjusted with the same F-contrast), or would you advise against it?
Yes - this is fine.
> Secondly: for the DCM analysis, I'd be interested in only including 4 of the 6 conditions in the design matrix (from my understanding the 2 remaining conditions would form the implicit baseline), and include all 4 together as driving input (a 'task' condition with cond1+cond2+cond3+cond4) and each separate condition as modulatory input.
Yes that sounds good.
> The subsequent contrasts in the PEB will be different from the ones I used in the GLM to extract the VOIs. Is this ok too? > I would just like to make sure that it is correct to have different number of conditions/contrasts in the GLM for VOI extraction vs in the DCM design/PEB analysis.
I suggest modelling as much as possible in the DCM (because the BOLD responses of neighbouring trials will overlap, and they can only be separated out if the model "knows" when your trials occurred). So don't have more than one unmodelled baseline condition. Then for the PEB, just take up to the group level the DCM parameters of interest. Typically these will be the B-matrix parameters (i.e., modulations) related to your conditions of interest.
Best [log in to unmask] |