Dear Marco,
Apologies for my delayed response, but life is just too busy at the moment.
As a precursor, the goal of the analysis in the papers by Heim et al., Human Brain Mapping, in press, and by Stephan et al. 2007, J. Biosci., is the same, i.e. to explain the occurrence of a two-way interaction in a particular region in terms of modulation of afferent connectivity.
I don't know what your exact SPM results look like, but here are two options how you could implement the same kind of analysis given that you have found a significant interaction (STIMULUS x TASK) in a region Y:
(a) If your SPM analysis shows a significant main effect of STIMULUS in region X, then you can test whether
you can explain the interaction in region Y by driving region X with stim1 and stim2 separately (or with a single input encoding their difference) and modulating the connection X-->Y by task1 and task2 separately (or by a single input encoding their difference).
(b) If your SPM analysis fails to show a significant main effect of STIMULUS anywhere, but you find a region X responding to all stimuli alike, then you can test whether you can explain the interaction in region Y by driving region X with a single input encoding all stimuli and modulating the connection X-->Y by an input encoding the STIM x TASK interaction.
In summary, note that the question is not necessarily about whether the connection is modulated by the interaction; this depends on the nature of the response in the source region. In contrast, the question is whether you can explain, in terms of task-dependent changes in connectivity, the occurrence of an
interaction in a target region.
Should anything still be unclear, have a second look at the two papers mentioned above; the rationale underlying the analysis is described in detail there.
Best wishes,
Klaas
----- Ursprüngliche Mail ----
Von: marco tettamanti <
[log in to unmask]>
An:
[log in to unmask]Gesendet: Dienstag, den 25. März 2008, 16:59:37 Uhr
Betreff: [SPM] DCM and 2x2 factorial design
Dear Klaas and SPMers,
I am trying to setup a convenient DCM model for a 2x2 factorial design,
and I would very much appreciate your advice to be sure that what I am
doing is correct.
The model is constituted by 4 anatomical regions, with only one region
receiving a driving input.
There are two different questions that I would like to address:
1. Is the overall functional integration within the
system stronger for
stimulus A vs B (main effect of stimulus (A1+A2)-(B1+B2))?
2. Is there a significant interaction between the two factors, such that
the connection strengths for stimulus A vs B are stronger in the context
of task 1 vs 2 (interaction (A1-B1)-(A2-B2))?
To address these two questions I have set up a GLM Design matrix with
two conditions:
- onset of events of condition A (representing the main effect
(A1+A2)-(B1+B2)) -> 'STIMULUS'
- onsent of events of condition 1 (representing the main effect
(A1+B1)-(A2+B2)) -> 'TASK'
I then defined a DCM model, in which 'STIMULUS' provides the driving
input, and 'TASK' is allowed to modulate all the defined intrinsic
connections.
Is the procedure described above so far correct?
Do I understand it correctly that:
- The intrinsic connection strengths (matrix DCM.A) represent how the
main effect of 'STIMULUS' that is induced in the
input area is conveyed,
via the specified connections, to other areas (answer to question 1 above).
- The modulatory connection strengths (matrix DCM.B) represent the
degree to which the activity is passed from the input region to other
areas, depending on the level of 'TASK' (answer to question 2 above).
Thank you for your help!
Best wishes,
Marco
--
Marco Tettamanti, Ph.D.
San Raffaele Scientific Institute
Facoltà di Psicologia
Via Olgettina 58
I-20132 Milano, Italy
Tel. ++39-02-26434888
Fax ++39-02-26434892
Email:
[log in to unmask]