Hi,
The run-time will scale pretty much linearly ( for a given data set ) with number of permutations and number of contrasts. The randomise_parallel script uses an (extremely) conservative estimate of 100 TFCE permutations in 30 minutes.
Many Regards
Matthew
> Hello,
>
> Can I calculate an approximation of time for randomise step? Is there any general formula?
>
> thanks,
> Alba Garin
> Simulation Area
> Applied Mechanics
> CEIT
> San Sebastian - Spain
> ________________________________________
> De: FSL - FMRIB's Software Library [[log in to unmask]] En nombre de Stephen Smith [[log in to unmask]]
> Enviado el: miércoles, 18 de enero de 2012 10:03
> Para: [log in to unmask]
> Asunto: Re: [FSL] timing estimation TBSS
>
> Hi - Tom recommends 5000 permutations, but 500 will often give very similar results, and is certainly good enough for an "initial" analysis.
> Cheers.
>
>
> On 16 Jan 2012, at 11:56, Garin, Alba wrote:
>
> Hi
>
> Should I set the number of permutations to 500 always? (even for a N=50 test, where I need more permutations for the statistics)
> How does this affect to my results?
>
> thanks in advance,
> Alba
> ________________________________________
> De: FSL - FMRIB's Software Library [[log in to unmask]] En nombre de Garin, Alba [[log in to unmask]]
> Enviado el: lunes, 16 de enero de 2012 11:54
> Para: [log in to unmask]<mailto:[log in to unmask]>
> Asunto: Re: [FSL] timing estimation TBSS
>
> Hi,
>
> but i need to do the all subjects to all subjects comparative.
> How can I calculate the amount of time that I need for a tbss run of N subjects? how many permutations are needed in the randomise command? should i thresholded to 500?
> I did a 13 subjects' comparative and i needed 1716 permutations. it increases to 140 min the time needed for this step. But when i did a 8 subjects' comparative I needed only 3 minutes for this step.
> why did it happen?
>
> thanks,
> Alba
> ________________________________________
> De: FSL - FMRIB's Software Library [[log in to unmask]] En nombre de Stephen Smith [[log in to unmask]]
> Enviado el: lunes, 16 de enero de 2012 11:28
> Para: [log in to unmask]<mailto:[log in to unmask]>
> Asunto: Re: [FSL] timing estimation TBSS
>
> Hi - if you are using the default recommendations for the registration, you now need in general just N of them, not N*N.
> Cheers.
>
>
> On 16 Jan 2012, at 09:45, Garin, Alba wrote:
>
> Hi everyone,
>
> I need a timing estimation for a study of tbss for a big amount of subject (N = 50, N=100, or even N=150), and I saw in www.fmrib.ox.ac.uk/fsl/tbss<http://www.fmrib.ox.ac.uk/fsl/tbss>/index.html<http://www.fmrib.ox.ac.uk/fsl/tbss/index.html> one formula that there is an approximation for the non linear registration (tbss_2_reg) for all-subjects-to-all-subjects options ( 5min x N x N). I did an experiment for N = 4, N = 8, N=13 using t test and this options for N=13:
> fsl4.1-tbss_1_preproc *.nii.gz
> fsl4.1-tbss_2_reg -n
> fsl4.1-tbss_3_postreg -S
> fsl4.1-tbss_4_prestats 0.2
> cd FA
> fsl4.imglob *_FA.*
> cd ../stats
> fsl4.1-design_ttest2 design 7 6
> fsl4.1-randomise -i all_FA_skeletonised.nii.gz -o tbss -m mean_FA_skeleton_mask.nii.gz -d design.mat -t design.con -n 500 --T2 -V
> fsl4.1-tbss_fill tbss_tfce_corrp_tstat1.nii.gz 0.95 mean_FA.nii.gz tbss_fill
>
> - could you give me any hint of formula for each step of TBSS, please?
> I know that it also depends on the computer you use, but I need any hint (mine is a Intel Core 2 Duo E8400 @ 3.00GHz 3.2 GB RAM that runs in ubuntu).
> - Is it possible to do tbss for a N=150 group of subjects?
>
> Thanks in advance,
> Regards,
>
> Alba Garin
> Simulation Area
> Applied Mechanics
> CEIT Research Centre
> San Sebastian - Spain
>
>
> ---------------------------------------------------------------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask]<mailto:[log in to unmask]> http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
>
>
>
> ---------------------------------------------------------------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask]<mailto:[log in to unmask]> http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
>
|