Hi,
I'm trying to prepare a go/nogo task for the scanner and I want to make sure my go's and nogo's are not collinear. At first, using the -nodata option in the following command and a TR of 2.5 seconds, I get a good result:
bash-3.2$ 3dDeconvolve -nodata 1098 2.5 -polort A -num_stimts 2 -stim_file 1 'go.txt' -stim_label 1 'GO' -stim_file 2 'go.txt' -stim_label 2 'NOGO'
However I am considering reducing the TR length, and wanted to see if they would remain collinear. I noticed that the output didn't change with a smaller TR, and then I tried drastically reducing the TR and even with 0.005s TRs the output is almost the same and tells me that the conditions are not collinear (see below). This seems impossible, since with an HRF peaking at ~5 seconds, and a full task length of 5.5s I cannot imagine that the go's and nogo's could be resolvable here.
How can I include some idea of HRF length (like GAM?) so that I can test if these are collinear and not just if the matrix itself is solvable?
Thanks!
Michael
bash-3.2$ 3dDeconvolve -nodata 1098 0.005 -polort A -num_stimts 2 -stim_file 1 'go.txt' -stim_label 1 'GO' -stim_file 2 'nogo.txt' -stim_label 2 'NOGO
'
++ 3dDeconvolve: AFNI version=AFNI_2011_12_21_1014 (Jul 30 2013) [64-bit]
++ Authored by: B. Douglas Ward, et al.
++ using NT=1098 time points for -nodata
++ Imaging duration=5.5 s; Automatic polort=1
++ Number of time points: 1098 (no censoring)
+ Number of parameters: 4 [2 baseline ; 2 signal]
++ Wrote matrix values to file nodata.xmat.1D
++ ----- Signal+Baseline matrix condition [X] (1098x4): 1.70702 ++ VERY GOOD ++
++ ----- Signal-only matrix condition [X] (1098x2): 1 ++ VERY GOOD ++
++ ----- Baseline-only matrix condition [X] (1098x2): 1 ++ VERY GOOD ++
++ ----- polort-only matrix condition [X] (1098x2): 1 ++ VERY GOOD ++
++ Wrote matrix values to file nodata_XtXinv.xmat.1D
++ +++++ Matrix inverse average error = 5.57179e-16 ++ VERY GOOD ++
++ Matrix setup time = 0.01 s
Stimulus: GO
h[ 0] norm. std. dev. = 0.0663
Stimulus: NOGO
h[ 0] norm. std. dev. = 0.0893