On Thu, Nov 24, 2011 at 9:18 PM, Alex Benin <[log in to unmask]> wrote:
> Hello SPMers! I'm trying to make my first DCM model for ERP and I'd like to clarify some moments:
> 1)My ECD model was done using ERP grand average data for more than 50 subjects, so can I apply this grand average data to SPM or I should choose only one person from my study?
You can fit a DCM-ERP to the grand average, but if for instance you
want to do classical statistical tests on model parameters or random
effects model comparison then you need to fit individually to each
subject. It is possible to combine the two approaches, i.e. fit a
series of models to the grand average, select the best model and then
fit it to each subject and do classical stats on the parameters.
> 2)My ECD model contains both single and regional dipoles though residual variance is rather good, so is it applicable for DCM study?
Regional sources is something that only exists in BESA as far as I
know. You can either use ECD or IMG option in DCM and put the sources
where your dipoles and regional sources are.
> 3)The quantity of dipoles is 8, so the model may be too complex - is it possible to throw away those sources, in which I'm not interested in? (to compare models with different connections)
For the model comparison to be valid you should always include all the
sources in all your models but you can exclude some sources by not
connecting them. The rule of thumb is the models you compare should
only differ in their connectivity matrices and input timing.
> Thank you very much!