Dear Clément
Sorry for the delay in replying. In case this is still helpful...
I suggest you do not go back to the first level DCMs using the second output from spm_dcm_peb. Rather, make all your inferences by comparing PEB models (spm_dcm_peb_bmc), and look at the parameter estimates that result. The outputs will include the probability for an effect of group on each connection. There's no need to do a t-test.
Best
Peter
-----Original Message-----
From: SPM (Statistical Parametric Mapping) <[log in to unmask]> On Behalf Of Clément Tarrano
Sent: Thursday, March 30, 2023 6:02 PM
To: [log in to unmask]
Subject: [SPM] Overall inhibitory/excitatory effect using DCM PEB
⚠ Caution: External sender
Dear DCM experts,
I would like to compare the effective connectivity at rest in a network containing a cerebellar region and several cortical nodes between patients and controls.
I am also interested to study the overall effect of the cerebellum on the cortex in term of excitation/inhibition and differences between the groups.
For this, I constructed a full model and used the PEB framework. This is my model matrix M : a first column of "1" and a second column with "1" for patients and "-1" for controls (then mean centered). I used the option "search over the nested model" to prune away non-relevant connections.
This analysis gave me two A-matrix contrasts: A first one for the entire group of participants and the second one for the between groups contrast.
Now, I would like to investigate the overall difference in term of inhibition of the cerebellum on the cortical nodes between groups.
I used [PEB,DCM ] = spm_dcm_peb(GCM,M) to get the subjects parameters updated with the group priors.
Then, I calculated the mean of the extrinsic connectivity parameters (with a posterior probability >99 for the entire group of participants - given by the first A-matrix contrast) directed from the cerebellum to the cortex in each groups.
Then I compared these means which were both negative, but more negative in controls compared to patients.
I have following questions :
- Can I simply assume from the above, that the cerebellum had a lower inhibitory effect on the cortex in the patients group ?
- Or do I need to perform a statistic test to compare these means (such as a t-test) or the update of the parameters made by the PEB has already taken into account the group difference?
- Is there a better way to resolve this question?
Thank you very much for your help.
Best,
Clément
|