> My plan was to run 3dDecon with all the event regressors (200 in my case)
> and just output the coefficients (not all the model stats). Extract the beta
> series from that output for my ROI and then use that ROI beta series to run
> 3dregana on the whole brain beta series (and convert the output to r stats) to
> find regions of functional connectivity. Is this what others are doing?

Are you talking about option -stim_times_IM in 3dDeconvovle? If so, it would be very interesting to see how well this whole thing works, and let us know how it goes if you don't mind.

> would the context-dependent correlation method described on Gang's AFNI page do something similar for me?

What you've described with "beta series" approach is looking for response similarity between the seed and the rest of the brain through individual events across time. Context-dependent correlation analysis (or PPI in the popularized name) is slightly different in the sense that it searches for the effect of interaction at "neural level" between the seed (physiological) and events (psychological) on a target region.

Gang