Dear SPM users,
I want to run a DCM analysis, but here are some problems.
In our experiment, we have five conditions (C1~C5).
The preliminary tradtional GLM (i.e., non-DCM analysis) showed that C2 condition, when contrasted against C1 and against C4, elicited higher activation in two areas (PFC and temporal pole).
The original GLM design matrix is attached.
Based on this result and previous literature, we hypthosized there is a temporal sequence between the two areas for C2 condition, so we want to run a DCM analysis.
To explore this modulatory effect of C2, I think I have to recombine the regressors I used in the original design matrix into two regressors, (as is done in the Chap.35 in SPM12 manual),regressor 1 = all condtions, regressor 2 = C2 only.
Then specify two t-contrasts respectively against baseline and an F-contrast for effects of interest.
The DCM matrix is attached.
My question is, why is it that, after I recombined the regressors, the activation in the new design matrix became extremely different.
That is, the C2-against-baseline contrast now reveals little activation across the brain (in the original traditional GLM, when I looked at the C2-against-baseline contrast, the activation map was broad and included the two critical regions).
Now from each subject I can barely find activation in temporal pole so I cannot extract time series signal from this region.
Any help will be greatly appreciated.
Thanks.
Kuan-Jung
|