Dear experts,
It would be very much appreciated if anyone could answer these burning questions. I was wondering whether the following is a valid way to create DCM models, as I want to assess whether the anterior insula has non-linear gating effects on DMN and CEN regions during the n-back task. (fMRI)
1. Are the non-linear coupling parameters in the D-matrix automatically estimated across the entire task (both 0-back and 2-back) or does this depend on how I write the script?
2. I have three models with 5 nodes:
Model 1: non-linear effects of anterior insula,
Model 2: non-linear effects of two CEN nodes together,
Model 3: non-linear effects of two DMN nodes together.
Is it valid to model two nodes of a network together, to have non-linear effects on the connectivity between other regions?
3. Besides having a fully connected intrinsic A matrix, I also want to use a fully connected B-matrix for the 2-back condition. After BMS and BMA, the estimations from the modulatory effects are then used to test for any differences in connectivity between patients and healthy controls. All three models thus have the same number of parameters and therefore the same degree of complexity. Is it justified to use a fully connected B-matrix in this sense? is there any risk of over-fitting?
4. Lastly, if I want to compare the modulatory effects between subjects, do I first need to assess their value relative to the A matrix values or can I just compare their absolute values with other subjects (using the same model of course).
Thank you!
With kind regards,
Felix
|