Posted by Adam Greenberg on March 26, 2002 at 20:01:00:
We are trying to apply 3dDeconvolve with the -nodata option to test some paradigms that we have. Because our actual stimulus only lasts 500 ms, we are interested in using the sub-TR option offered by -stim_nptr. From the decription in the manual, we should multiply the actual lag in TR units by nptr to specify -maxlag correctly.
We have a TR of 2500, and are using -stim_nptr k 5. We would like to estimate the initial 15 seconds, or TRs [0..6]. To do this we employed maxlag of 30 (6 * 5). To start out, we chose the simplest test just to check whether we had things set out correctly and employed a glt matrix such as:
0 0 1 1 1...(31 times)
I then ran the following command:
3dDeconvolve -nodata -nlast 739 -num_stimts 1 -stim_file 1 'A.txt' \
-stim_maxlag 1 30 -stim_nptr 1 5 -glt 1 A.mat
and received the following error message:
3dDeconvolve Error: Improper X matrix (cannot invert X'X).
Presumably we still have a mismatch between what 3dDeconvolve expects and the parameters that we used.
Thanks,
Adam