Dear Amit Etkin
Since I don't see you got any response till now, I can try to answer some
of your questions from my (user) understanding of DCM.
1- As for the design of the model:
In your model the direct input (C) is perfectly correlated with the
modulatory effects (B), which leads to the underestimation of connections
in the model, especially the connections with the input regions (both
regions in your case). You should include another condition in the model
that encompases both conditions 1 and 2 (and possibly other conditions in
you original analysis of 7 conditions, that exert their direct input on
the same region). This new condition will be assigned the direct input and
no modulatory effects, while conditions a and b will not have a direct
input. This way your direct input and modulatory effects are orthogonal.
2- The intrinsic connections are the strength of connections between
regions across the entire time series (all 7 conditions).
3- The modulatory effects are the strength of connection relative to the
intrinsic connections, and not a direct comparison between conditions 1
and 2.
4- The units in DCM are arbitrary, so they are only meaningful within the
given DCM model.
5- If you have identical models for all subjects you can take the A, B,
and C values into a second level RFX analysis outside SPM (t-tests, ANOVA
etc.)
I hope this was helpful.
Tali Bitan
Northwestern University
On Thu, 20 Apr 2006 03:37:55 +0100, Amit Etkin <[log in to unmask]> wrote:
>Hello all!
>
>This is a general question about DCM setup and interpretation that relates
>to my data, but I hope will have relevance to many other users.
>
>I have two regions of interest from an fMRI data set which I found to be
>activated by interesting contrasts in a task, and are coupled to each
other,
>as determined by a PPI analysis (comparing condition 1 coupling to
condition
>2 coupling). What I would like to understand is how condition 1 vs
condition
>2 affects the region A to B path separate of the B to A path. This will
>hopefully give some directionality to the PPI effect across the two
>conditions. I am trying to set up a DCM model and have several questions.
>Please tell me if any of the below are right or wrong:
>
>1. For intrinsic connections I chose A to B and B to A.
>2. For which conditions to include (out of 7) I chose conditions 1 and 2.
>3. For effects of condition 1, I chose it to affect region A, region B,
the
>path from A to B and the path from B to A.
>4. same as above for condition 2
>
>from what I understand, the area of the results I should be looking at is:
>DCM.A (intrinsic connectivity)
> -1.0000 0.0252
> -0.0193 -1.0000
>
>DCM.B (effects of conditions 1 and 2) on A to B and B to A paths
>ans(:,:,1) =
> 1.0e-004 *
> 0 0.2279
> -0.0309 0
>
>ans(:,:,2) =
> 0 0.0003
> 0.0016 0
>
>DCM.C (input effects)
> 0.0007 -0.0684
> 0.0022 -0.0264
>
>
>Questions:
>1. What do the intrinsic connection values mean? and are they in relation
to
>the conditions in any way?
>
>2. Am I right in interpreting the DCM.B results as saying that, relative
to
>condition 1, condition 2 led to a decrease in B to A path strength and an
>increase in A to B path strength?
>
>3. Am I right in saying that, relative to condition 1, condition 2 led to
>less activation in regions A and B (values going from slightly positive to
>negative)?
>
>4. Should I include all 7 conditions? If so, how should I set up the model
>if all I'm interested in is the effects of conditions 1 and 2 in A to B
and
>B to A paths.
>
>5. Can I take the values for each of these items (once I understand them)
to
>the group level by a simple t-test approach?
>
>6. What unit of measure do the numbers in DCM.A, .B, and .C refer to?
>
>much thanks!
>
>Amit
>=========================================================================
|