Questions
1) Given an event ("mem") repeated in each run, can the following be interpreted as the same:
a) modeling all events across runs using a single -stim_times ( -stim_file "all_runs.1D" -stim_label "mem")
b) modeling each run separately (-stime_label "mem_run1" ... -stim_label "mem_run2") and combining with gltsym ("gltsym:mem_run1 + mem_run2 + mem_run3" -glt_label "mem")

2) Is using 3dDeconvolve with per-run stim_files an odd thing to do? Would it be more rigorous/reasonable to use 3ddeconvolve on each run independently and use other tools to build the contrast -- especially if the contrast will be used with covariates (e.g. age, group, behavioral scores)?

This is more obvious when mem_run1 = memorizing faces, mem_run2 = memorizing cars, and the desired contrast is faces - cars


Thanks!!

Context

We have 3 runs of an image memorization+recall task. each run used a different set of images: 2 different sets of faces (AUS, USA) and one set of cars. The task was performed by participants in one group (ASD) and in another of matched controls.

Initially, we combined the 3 runs modeling memory+{correct+error}_recall events to contrast "correct recall" and "mem" with gltsym.

  -stim_times_AM1 1  ${timebase}/Mem.1d 'dmBLOCK' -stim_label 1 'mem' \
  -stim_times_AM1 2  ${timebase}/Test_crct.1d 'dmBLOCK' -stim_label 2 'corr' \
  -stim_times_AM1 3  ${timebase}/Test_err.1d 'dmBLOCK' -stim_label 3 'err' \
...
  -gltsym "SYM:corr -mem"  -glt_label 1 'corr-mem' \

With a 1D timing file (Mem.1D) like

15.00:9.00 51.00:9.00 99.00:9.00 148.50:9.00 186.00:9.00 231.00:9.00
15.00:9.00 57.00:9.00 100.50:9.00 150.00:9.00 190.50:9.00 231.00:9.00
10.50:9.00 55.50:9.00 93.00:9.00 136.50:9.00 183.00:9.00 228.00:9.00

But that does not distinguish between face and car events. To contrast events between the two, I created 1Ds specific to each run, like "Cars/Mem.1d", with empty rows where the other runs are

-1:0
15.00:9.00 57.00:9.00 100.50:9.00 150.00:9.00 190.50:9.00 231.00:9.00
-1:0

and a call to 3dDeconvolve that looks like

         -stim_times_AM1 7  ${timebase}/AUS/Mem.1d        'dmBLOCK' -stim_label 7  'Amem' \
         -stim_times_AM1 8  ${timebase}/AUS/Test_crct.1d  'dmBLOCK' -stim_label 8  'Acorr'\
         -stim_times_AM1 9  ${timebase}/AUS/Test_err.1d   'dmBLOCK' -stim_label 9  'Aerr' \
         -stim_times_AM1 10 ${timebase}/Cars/Mem.1d       'dmBLOCK' -stim_label 10 'Cmem' \
         -stim_times_AM1 11 ${timebase}/Cars/Test_crct.1d 'dmBLOCK' -stim_label 11 'Ccorr'\
         -stim_times_AM1 12 ${timebase}/Cars/Test_err.1d  'dmBLOCK' -stim_label 12 'Cerr' \
         ... \
         -gltsym "SYM:Amem +Cmem +Umem"     -glt_label 1 'mem' \
         .... \
         -gltsym "SYM:.5*Acorr +.5*Ucorr -Ccorr"  -glt_label 7 'face-car_corr'