Dear Elisa,
The same thinking applies here whether it’s a GLM or a DCM.
A positive parameter estimate corresponds to connections that show a positive difference for load 1 (c1) versus load 2 (c2).
Here c1 is the change in connectivity parameter for load 1.
Here c2 is the change in connectivity parameter for load 2.
This difference could arise for
1. c1 > c2 (both neg)
2. c1 > c2 (both pos)
3. c1 > c2 (c1 pos, c2 neg)
4. c1 > c2 (c1 zero, c2 neg)
5. c1 > c2 (c1 pos, c2 zero)
There's no way of telling these apart unless you have separate modulatory regressors for c1 and c2. This requires you to enter c1 and c2 as separate conditions in your original design.
But maybe your question is answered with just c1 > c2.
Best,
Will.
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
> On Behalf Of Elisa Scheller
> Sent: 05 March 2013 10:57
> To: [log in to unmask]
> Subject: [SPM] Interpretation of parametric modulator in DCM B Matrix
>
> Sorry, the message text of my previous post somehow got lost in the
> attachment. Next try comes here:
>
> Dear DCMers,
>
> we have an n-back paradigm with two levels of working memory load (i.e.
> two of the three conditions in our blocked design). We'd like to enter
> a parametric modulator of these two conditions as an experimental
> modulation in the B Matrix.
> The respective DCM modulatory input is attached. It is probably just
> mean scaled (we entered 1 and 2 as pmod.param), but we are unsure how
> to interpret e.g. positive effects of this input function in the
> estimated DCM. Does a positive parameter estimate signify an increasing
> positive coupling with increasing working memory load (as intended)? It
> could also mean a decreasing coupling with low working memory load and
> a positive coupling with high load. If the first interpretation is
> correct, how will intervals with the input being zero (see picture) be
> interpreted? We would like DCM to test for changes in coupling only for
> the non-zero sections and specifically do not want them to be
> considered as an intermediate working memory load (between the positive
> and the negative values).
>
> It would be great if someone could help us to understand why the
> conditions are depicted like that and how resulting positive/negative
> parameter estimates should be interpreted.
>
> Many thanks,
> Elisa
|