Hello Andrew,
The /PPHeights values are not used by randomise. You shouldn't necessarily need to queue randomise_parallel itself ( the actual script will run quickly and generate/submit two further scripts ). One way to test the model is to just run the normal randomise binary with the same inputs but only a small number of permutations ( e.g -n 4 ). One other thing to check is the *.o and *.e log files for the individual randomise runs - they may contain information about what has failed..
Many Regards
Matthew
> Dear FSL Gurus,
>
> I'm a FSL novice and am attempting to use Randomise_parallel to run a TBSS model, but it seems that the Seed defragmentation is not working and the process will run for as long as I let it (weeks, even).
>
> I'm running the latest version of FSL on a Mac Pro with 64GB RAM and dual 12-core processors. However, instead of SGE, I am using Mac OS XGRID with a modified FSL_SUB bash file <https://rcbi.rochester.edu/users/vanooste/weblog/f657b/fsl_sub_implemented_for_local_multiprocessor_systems.html> which could be a contributing factor. For example, that modified script seems to break bedpostx from running correctly.
>
> So, I'm not sure if the problem is that the model is not set up correctly, or if it's due to some other issue with the parallel processing component.
>
> As far as the model is concerned, I would greatly appreciate some assistance with reality checking. There are 118 subjects in 2 groups. I'm trying to get Randomise to model group differences as well as age as a regressor (already demeaned) and the interactions thereof. In addition, I'm unclear what the /PPheights command does in randomize for DW data and how to determine those values for the design.con and design.mat files. The documentation that I have found has only added to my confusion.
>
> Randomise_parallel runs and generates 13 SEEDS rather quickly but doesn't seem to run the defragment. I read through some previous posts to the email list (link here: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;b9d28ec9.1003) that suggested making a couple of changes to the randomise_parallel script (e.g., changing #!/bin/sh to #!/bin/bash), and changing the fsl_sub command to veryshort.q instead of the default. However, because neither solution seems to address the problem, I'm not sure if it's a problem with the randomize model or due to the modified fsl_sub_bash file.
>
> Below is the command syntax and design.mat and design.con files I've used:
>
> Command: fsl_sub -q veryshort.q randomise_parallel -i /MY_FOLDER/mytbss/stats/all_FA_skeletonised -o tbss -m mean_FA_skeleton_mask -d design.mat -t design.con -n 500 -c 3 -V --T2
>
> design.mat:
> /NumWaves 5
> /NumPoints 118
> /PPheights 1 1
> /Matrix
> 0.00 1.00 -29.12 -29.12 0.00
> 0.00 1.00 -12.12 -12.12 0.00
> 0.00 1.00 5.88 5.88 0.00
> 0.00 1.00 -0.12 -0.12 0.00
> 0.00 1.00 2.88 2.88 0.00
> 0.00 1.00 -2.12 -2.12 0.00
> 0.00 1.00 31.88 31.88 0.00
> 0.00 1.00 -2.12 -2.12 0.00
> 0.00 1.00 1.88 1.88 0.00
> 0.00 1.00 14.88 14.88 0.00
> 0.00 1.00 10.88 10.88 0.00
> 0.00 1.00 4.88 4.88 0.00
> 0.00 1.00 9.88 9.88 0.00
> .
> .
> .
> 1.00 0.00 26.88 0.00 26.88
> 1.00 0.00 -13.12 0.00 -13.12
> 1.00 0.00 -13.12 0.00 -13.12
> 1.00 0.00 -8.12 0.00 -8.12
> 1.00 0.00 7.88 0.00 7.88
> 1.00 0.00 -19.12 0.00 -19.12
> 1.00 0.00 -32.12 0.00 -32.12
> 1.00 0.00 -26.12 0.00 -26.12
> 1.00 0.00 -28.12 0.00 -28.12
>
> design.con:
> /NumWaves 5
> /NumContrasts 5
> /PPheights 1 1 1 1 1
> /Matrix
> 1 0 0 0 0
> 0 1 0 0 0
> 0 0 1 0 0
> 0 0 -1 0 0
> 0 0 0 1 -1
> 0 0 0 -1 1
>
> Thank you for your expertise, patience, and assistance!
>
> Andrew
>
> ********
> Andrew R. Bender, M.A.
> Graduate Research Assistant
> Cognitive Neuroscience of Aging Lab
> Department of Psychology
> Institute of Gerontology
> Wayne State University
> Detroit, Michigan 48202
> lab 313.664.2649
> fax 313.664.2667
>
|