Dear DCM experts,
I am interested in how the selection of the input region may influence the
outcome in a DCM. Say I have two regions, A & B, responding to task 1 > task
2. I have a third region, X, that responds to (task 1 + task 2) > baseline
and is in a plausible primary sensory area. I compare two models to test
whether regions A & B interact or are independently responding to a common
input. Both models have direct inputs from both task 1 and task to to X and
intrinsic connections between all regions. Modulating effects of task 2 only
are shown below:
Model 1
/ A
X
\ B
Model 2
A
X |
B
My question is, if I find that model 1 gives better fit than model 2, does
this mean my regions do not interact, or does it simply mean that modulating
effects are stronger when acting on connections closer to the input? I ask
following Klaas' comment about effects propagating through the network,
becoming weaker as they get further from the input.
https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind05&L=SPM&P=R581085
As an aside, I have been using a separate input region because I have an
event-related design, and this region "mixes" the responses to all stimuli,
as recommended by Karl here:
https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind04&L=SPM&P=R237524
An alternative would be use a new design matrix with an explicit "all
stimuli" column, and include its direct input on all task regions. This
would force activation to be explained by interactions between task regions,
so I would be happier doing this if I could justify missing out a specific
input region.
Many thanks for your help,
Paul Wright
|