Hello Gaurav,
Can you upload the design that fails, and another design that works well to:
https://oxfile.ox.ac.uk/oxfile/work/extBox?id=700839B35190E48BF
also can you let us know the voxel dimensions of the task data file?
Kind Regards
Matthew
> I am using fsl_glm to regress out spurious noise form a resting state data set, and randomly am seeing a hard crash of fsl_glm that I was hoping someone could help me with. The data are from the HCP processing pipeline, and have been converted from cifti to nifti using wb_command. Since I am using the HCP pipeline, the FSL version is 5.0.6. The design matrix is 1064 time points x 228 regressors (there are lots of spike regressors for dvars and rms regression), and all duplicate regressors have been removed (those had been causing a similar crash earlier). The command I use is this:
>
> fsl_glm -i Task_fMRI_1_Atlas_x4_bpss.nii.gz -d Task_fMRI_1_x4_bpss_nuissance_regressors.dat -o betas.nii.gz --out_res=resid.nii.gz --demean
>
> and the crash I get is this:
>
> libc++abi.dylib: terminating with uncaught exception of type NEWMAT::SingularException
> Abort
>
> I am running this on a Mac Pro with 128gb RAM running OS 10.9.5. This exact command works well for shorter runs, so is it just the run length and/or the number of regressors? Any help would be appreciated, and thanks!
>
> Gaurav
|