Dear SPM experts,

I have a methodology question so please bare with me. 

I ran an ICA analysis on resting state data and I found that for one component there was an area in the brain that was more strongly connected in one group than the other: call that region R.

The next question I wanted to ask was: what is the connection profile through DCM between R and subregions of the component.

What to do next was a bit complicated. Subdividing a component into subcomponents is not so straightforward, but the advice I received and which I believe is correct was to create masks of subregions with the component and insert them in dual regression along with maps of the other components.

That gives us a timecourse for each of the subcomponents that can be inserted into a DCM analysis.

The next thing however was to find the component timecourse in R. What I did is multiply the spatial map of the component in that region with the overall timecourse outputted by the first ICA analysis.

I then inserted the timecourses of the subcomponents as well as the timecourse for that region of the brain (specific to the component) into a DCM analysis. The results made 'sense': however, I am worried that the timecourses for the subregions and the one for R have been derived in a different way, which may make the analysis invalid.

What do you guys think and any ideas how to proceed?