Hi all,
I am trying to run analyse a fairly large (816 timepoints) dataset.
I have 20 EVS, plus temporal derivatives, corresponding to several classes of event regressors, two error regressors and three block regressors (no temp. derivative included). I'm using 3 column format EVs with the double-gamma HRF convolution. I'm also orthogonalizing WRT motion.
The processor is a Mac 2 x 2.66 GHz Dual-Core Intel Xeon, with 8GB RAM (and oodles of free space). OS X Server 10.4.9
It crashes as follows:
/Users/Shared/fsl/bin/contrast_mgr -f design.fts stats design.con
contrast_mgr(2884) malloc: *** vm_allocate(size=2654208000) failed (error code=3)
contrast_mgr(2884) malloc: *** error: can't allocate region
contrast_mgr(2884) malloc: *** set a breakpoint in szone_error to debug
Uncaught exception!
Rendering using zmin=2.3 zmax=8
I tried it without the prewhitening step, I've also tried removing some regressors (errors) but can't really drop any more. Neither of these options worked. Because of the nature of the experiment, I can't analyse the data separately for each block then combine in a fixed effects analysis.
I do have a large number of contrasts (40) - does this contribute to the memory problem (i.e. could be solved if I ran half the contrasts in one feat and the other half in another)? Or is it purely a result of the large number of timepoints and large number of EVs? Perhaps I should just drop the temporal derivatives? I come across this memory problem frequently with FEAT and would really like to know how other people come to a solution.
Any advice would be greatly appreciated!
Cheers,
Clare
