Dear John,
I would not recommend the procedure that you suggest. Differences in the rate constants of connections (and their modulations) from different regions can be difficult to interpret unless there are no general differences in signal magnitude (e.g. when dealing with homotopic regions in the two hemispheres).
I would rather compare three models using BMS: one with the modulation on one connection, a second with the modulation on the second connection, and a third model with modulations of both connections.
Best wishes,
Klaas
________________________________
Von: John Herrington <[log in to unmask]>
An: [log in to unmask]
Gesendet: Dienstag, den 22. September 2009, 16:22:57 Uhr
Betreff: [SPM] Using DCM modulatory effects as descriptive statistic for GLM
I have a question about whether a particular way of using output from DCM is
valid. In particular, I was wondering if it was valid to use the
"modulatory effects" values of a single DCM as descriptive statistics that
could be submitted to other statistical inference tests (particularly
GLM-based ones). Say, for example, that I have a single DCM model that
contains only two brain regions (A and B), and only one task condition
(i.e., no condition-specific effects are being examined for present
purposes). The DCM is set up to allow for both bi-directional intrinsic and
modulatory effects between A and B. I want to ask a simple question - does
activity from A modulate B, or vice versa. The null hypothesis would be
that there's no difference between the modulatory effects of A and B on one
another. Could I treat the modulatory effect values of A-to-B and B-to-A as
descriptive statistics and submit them to a t-test to see if they differ? I
realize that this could be handled using the DCM comparisons approach (i.e.,
examinations of AIC/BIC) of two unidirectional models, but there seem to me
to be some advantages to using a null hypothesis test approach as above, it
it's feasible.
Thanks,
John
