Dear Mishna,

There can be various approaches to selecting timeseries for your DCM such that they are well-standardised across subjects.  The fundamental requirement is that across subjects timeseries are chosen such that they reflect the same experimental effect or functional profile (as assessed, for example, by an initial GM).  That is, the same contrast at the same threshold should be chosen for selecting timeseries in each individual subject. Often, however, you will want to add additional anatomical constraints.  For example, you could take the maximum from the group GLM and use this as a starting point for identifying the strongest local maximum, within each subject, that is located within a radius of  X millimetres around the group maximum and that is located within the same anatomically defined structure (e.g. a particular gyrus or sulcus).  You can finesse this approach by using probabilistic atlases, e.g. the SPM Anatomy Toolbox.

The paper by Heim et al. that Qing mentioned in her previous e-mail has adopted that approach.  You can also have a look at Stephan et al. 2007, J Neurosci, where this approach is described in full detail in the supplementary materials. See http://www.jneurosci.org/cgi/content/full/27/13/3512

Best wishes,
Klaas