Dear DCM experts,
I'm wondering if it's possible to use DCM to test if an influence
between two regions is direct or indirect. I noticed that in the Zeki
Lab's DCM Powerpoint presentation (available on the SPM website) they
mention that this is not possible.
It is not clear to me why running a model with A->B->C and A->C
connections wouldn't resolve this issue. If the A->C connection has a
high p-value then the model would be supporting the direct connection,
while high A->B and B->C p-values would support the indirect
connection.
Alternatively, wouldn't it be possible to run two models, one
including A->B->C and another simply including A->C, and then running
a model comparison procedure to determine which model is better?
Have any simulations been run to test if either of these approaches
would work? (Or is there some clear a priori reason why these couldn't
work?)
Thanks,
Mike
|