Connection strength that enters the equations in the model is actually
E*exp(A+c*B) where E is some physiologically meaningful prior value
and c is the number from 'experimental effects' box corresponding to
the condition being modelled. Also note that the values in A and B are
optimised by the fitting scheme, which means that changing c will
affect the values in B in such a way that A+c*B will not change
(assuming there is only one global maximum).
On Wed, Aug 17, 2011 at 9:50 PM, Xing Tian <[log in to unmask]> wrote:
> I am a novice of running DCM in SPM. I am doing the DCM ERP.
> Here are my 3 questions:
> 1. In the box of "specification of effects" for between trial effects, can I input [0 -2]? What is it different from input [0 -1]?
> The intrinsic (baseline) connection strength (coupling (A) in the results) will be evaluated using the response corresponding to the "0", right?
No. A is fitted based on all the conditions as A enters the equation
above for all the conditions. Just when c is non-zero it is combined
with B and when c==0 A enters alone.
> What if I have 3 responses and specify the between trial effects as [1 0 -1] to test the linear effect, what will be used to evaluate the intrinsic connection? the average of these 3 responses?
> 2. The coupling (B) in the results represent the posterior means and their probability of gain in all connections for modulation effects (fitting the difference between the responses in experimental condition and in baseline condition), right? Are the posterior means for the gain modulations always positive numbers?
It depends where you look. The values in DCM.Ep.B can be positive or
negative. But what is displayed in the GUI table is exp(DCM.Ep.B)
which is greater than 1 when connection strength increases and less
than 1 when it decreases.
> If so, where I can find the direction of modulation effects?
> Does it imply that the connection strength increases if the between trial effects are specified as [0 1] and decreases if specified as [0 -1]?
If you change the value of c the value of B will adjust and the net
result will be the same. But the interpretation of B depends on the
value of c.
> 3. How can I batch DCM analysis for multiple participants/models in SPM? Where I can find it and what procedures I should follow?
There is an example script man\example_scripts\DCM_ERP_subject1.m that
you can use as a basis for your script to configure DCM in code.
Another way is to use the attached function that will be included in
MEEGTools toolbox in the next SPM update. To use it you should specify
all your different models for one subject using GUI and save them in
separate files. The function will ask you to specify these DCM files
and a group of subject datasets and will run all the DCMs on all the
subjects. If you want the function to run uninterrupted you should
specify the head models for all the subjects. There is batch tool you
can use to do that before you run the DCM batch.