Hi,
Hernando is correct about the seed - you need to use an equals sign and not space.
As for the other issue, I am not 100% sure but I suspect it is related to the fact that your design matrix is rank deficient when you invoke the -D demeaning option. This is because your first two columns add up to a constant column (of all ones) which is fine, until you demean the design matrix and then they add up to zero which is rank deficient. We have tried to make our tools work with rank deficient designs as much as possible, which is probably why it is running for lots of permutations, but at some point it is easy to generate a divide by zero or other numerical error, which may cause problems in our code or in library code that we call.
So I suggest that you remove the -D option and see if that fixes it. The two contrasts that you include here are fine without demeaning, though in general what you should do it to demean the contents of EV3 and EV4 (though EV4 may already be zero, I didn't check) prior to entering them into the design matrix. This would be the best solution in general (though for your particular contrasts at the moment it will not matter).
All the best,
Mark
> On 4 Aug 2017, at 21:24, Hernando Rafael Fereira <[log in to unmask]> wrote:
>
> Dear Jon,
> I am not sure about the first part of your error but as for your seed issue parameters set with a double dash need an equal sign to be set, e.g. --seed=9
> Best,
> Hernando
> ________________________________________
> From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Koller, Jonathan <[log in to unmask]>
> Sent: Friday, August 4, 2017 2:56 PM
> To: [log in to unmask]
> Subject: [FSL] randomise segmentation fault after specific permutation
>
> Hello FSL community,
>
> I am trying to use randomise for a group comparison. There are 19 images in group 1 and 14 images in group 2. I've pasted the contents of the design matrix (with age and sex in columns 3 and 4) and the contrasts below.
>
> The issue is, for some numbers of permutations, the program quits with a segmentation fault after a certain permutation. In the example output below, when using -n 10000, it gets to permutation 634, then crashes. It crashes at the same permutation again if re-running with -n 10000, but it might be, say, permutation 345 if we use -n 1000, or it might even complete if we use -n 5000. Since randomise isn't saving any output until it successfully completes all permutations for a contrast, we're having trouble tracking down the problem. Does anyone have any thoughts?
>
> If it helps, we seem to be able to always successfully run with -n 100, and we've successfully run with -n 1000 and once with -n 5000, too.
>
> P.S. We also tried using the --seed option, but couldn't figure out how to use it. E.g. --seed 1 and --seed 9999 both exit immediately with a " --seed: Missing non-optional argument!" error.
>
> Thanks,
> Jon
>
>
> randomise \
> -i HC_T1DM_dv_2year_c1_c2_masked_mni152_gauss_2mm.nii.gz \
> -o HC_T1DM_dv_2year_randomise_twopass_n10000 \
> -d HC_T1DM_FSL_randomise_design.mat \
> -t grp_comparison_FSL_randomise_contrasts.con \
> -m WOLF_TRACK_3tp_c1_c2_mask_avg_mni152_thr0p1.nii \
> -n 10000 \
> -T -D -R -P -N --twopass
> randomise options: -i HC_T1DM_dv_2year_c1_c2_masked_mni152_gauss_2mm.nii.gz -o HC_T1DM_dv_2year_randomise_twopass_n10000 -d HC_T1DM_FSL_randomise_design.mat -t grp_comparison_FSL_randomise_contrasts.con -m WOLF_TRACK_3tp_c1_c2_mask_avg_mni152_thr0p1.nii -n 10000 -T -D -R -P -N --twopass Loading Data:
> Data loaded
> Warning: Data demeaning selected, but at least one design column has non-zero mean - therefore invoking automatic demeaning of design matrix
> 8.18809e+08 permutations required for exhaustive test of t-test 1 Doing 10000 random permutations Starting permutation 1 (Unpermuted data) Starting permutation 2 Starting permutation 3 ... Starting permutation 633 Starting permutation 634 Segmentation fault (core dumped)
>
>
> cat HC_T1DM_FSL_randomise_design.mat
> /NumWaves 4
> /NumPoints 33
> /Matrix
> 1 0 19.4666 -1
> 1 0 15.5736 -1
> 1 0 17.1625 1
> 0 1 14.0889 -1
> 1 0 18.7764 -1
> 0 1 14.0861 1
> 1 0 13.7028 1
> 0 1 18.8736 1
> 0 1 12.9736 1
> 1 0 12.9736 1
> 1 0 16.8430 -1
> 1 0 7.9903 -1
> 1 0 8.8584 1
> 1 0 16.5139 -1
> 1 0 16.4042 -1
> 1 0 11.2903 -1
> 1 0 7.5278 1
> 1 0 12.8500 1
> 1 0 7.5972 -1
> 0 1 12.8667 1
> 0 1 12.1806 1
> 1 0 27.1639 -1
> 0 1 17.2472 1
> 1 0 24.1056 1
> 0 1 22.6819 1
> 0 1 14.5069 -1
> 0 1 8.4736 1
> 0 1 14.1375 -1
> 0 1 27.3097 -1
> 0 1 10.9528 1
> 0 1 21.1083 -1
> 1 0 18.1945 -1
> 1 0 17.0541 1
>
> cat grp_comparison_FSL_randomise_contrasts.con
> /NumWaves 4
> /NumPoints 2
> /Matrix
> 1 -1 0 0
> -1 1 0 0
>
>
> ________________________________
> The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.
|