Dear DCM-experts,
after humbly following the official advice to sit down and specify a
plausible anatomical model on a piece of paper before embarking on DCM
analyses, I have encountered some problems I haven't been able to solve:
I have a 2x4 factorial design, and I am interested in the interaction
between three regions (say A, B, C). The two level factor specifies a
driving input on region A, and the four level factor should modulate
connectivity between B and C. In addition, the connection between A and C
should be subject to an interaction between both factors. Here are my
questions:
- given that I assume a linear effect of the second factor on the B-C
connection, is it ok to specify the different levels of this factor as
eight seperate regressors and use a common contrast (i.e. repmat([-1.5,
-.5 .5 1.5],1,2)? Or is it better specify a model in which the different
levels are captured by a parametric modulation?
- regarding the interaction, I would like to demonstrate that the A-C
connection is only (or stronger) modulated by the four level factor for
one level of the two level factor. It seems to me that I could choose
between two alternatives: (i) defining the effects of the four level
factor separately for both levels of the other factor, let both effects
modulate the A-C connection and then use a paired t-test to compare the
parameters or (ii) specify two models: one in which the interaction (i.e.
[-1.5, -.5 .5 1.5 1.5 .5 -.5 -1.5]) modulates A-C and one in which this
modulation is absent. Comparing the Bayes factors of both models should
then identify the superior model. Any thoughts on what approach is more
appropriate?
Any hints would be greatly appreciated!
Best,
Thomas
--
---------------------------------------------------------
Dr. rer. nat. Thomas Wolbers
Department of Psychology
University of California
Santa Barbara, CA 93106-9660
U.S.A.
-
email: [log in to unmask]
www.psych.ucsb.edu/~wolbers
----------------------------------------------------------
|