Dear Boris,

DCM results are not independent from GLM results in the sense that both are generative models that are trying to explain the same thing, i.e. a measured BOLD timeseries. However, they have different mechanistic formulations and, depending on the model, it may not always be easy to identify relations between specific GLM parameters and specific DCM parameters. In your particular case, it is difficult to say anything about the nature of your modulatory effects without knowing the model in detail. Their interpretation will depend on how you modelled the activity in the region from which the connections, which you allowed to be modulated, originated.

Have a look at the recent e-mail on a similar topic which I sent to Marco. Does this help in re-specifying your model?

Best wishes,
Klaas

----- Ursprüngliche Mail ----
Von: boris suchan <[log in to unmask]>
An: [log in to unmask]
Gesendet: Donnerstag, den 24. April 2008, 11:05:36 Uhr
Betreff: [SPM] Relation DCM /GLM or signal changes

Hi all,

This might be a strange question, I controlled everything twice but….

We did an experiment with a 2 x 2 factorial design (Factor A (A1 A2) Factor B (B1 B2).

In the “normal analysis” we found a significant interaction resulting in a frontal and posterior activation.

To explore this interaction further, we extracted the signal changes which showed high signal changes for A1 and B2 compared to low signal changes in A2 B1.

In the next step we use DCM to look whether it is a top down / bottom up regulation.

Preliminary results of the DCM analysis show higher modulatory effects for condition A2 or B1 but not for that found in the interaction (A1/B2).

Can this be the case? Or is DCM and its results more or less independent from results shown in a GLM or with signal changes.

Many thanks

boris