Is there a statistical advantage/disadvantage over using -gltsym in 3dDeconvolve to average across conditions in a factorial experiment vs. creating separate .1D files collapsing across conditions?
Example:
2 x 2 x 2 within subjects design with (only) eight trial per 8 crossed conditions - total of 64 ER trials.
To test the overall effect across all conditions (there is also a between subjects group variable) one could have a separate .1D file for each 8 conditions and:
1) use -glt_label 1 Average_Response
-gltsym 'SYM: +0.125*cond1 +0.125*cond2 ... +0.125*cond8'
or
2) create a .1D file with the time points for all 64 trials and use -stim_times 1 Average.1D 'GAM' or the like in 3dDeconvolve.
Which (if either) of these is statistically more powerful?
p.s. thanks again to Rick and Ziad for the AFNI workshop last week in Wisconsin!!