Hi -

I have a condition A that can be decomposed into sub-conditions a_1 and a_2. Contrasting the sub-conditions with each other is of experimental interest, so right now the stimuli for a_1 and a_2 are specified directly to 3dDeconvolve in something like the following (simplified for illustration):

   3dDeconvolve -input pb04.${subjDay}.r*.scale+tlrc.HEAD                      
      -stim_times_AM2 1 a_1.1D "BLOCK(4, 1)"
      -stim_label 1 a_1
      -stim_times_AM2 2 a_2.1D "BLOCK(4, 1)"
      -stim_label 2 a_2
	  [...]
      -gltsym 'SYM: a_1[0] -a_2[0]'                                             
      -glt_label 1 'a1-a2'

That's all fine -- I get regressors for a_1 and a_2 and all is well. However, now I'd like to get an overall regressor that encapsulates both of these subtypes. I can't just add a new -stim_times that includes the union of a_1.1D and a_2.1D because that would be collinear with the existing regressors. I could re-run the whole analysis, replacing the two -stim_times with a single -stim_times that is the union of all a_1 and a_2 events, which would produce a 'parallel dataset' containing a single sub-brick for the superset A, but it seems like there would be a better solution than that. Is there?