| Author |
 Topic  |
|
vinod
Average Member
USA
68 Posts |
 Posted - Jul 30 2007 : 5:00:12 PM
|
| Then it looks likely that there are some other errors in your lower level analysis. Can you look at the report.log file in the same folder and look for errors? Also is there a stats folder created? If even the stats folder is not created, then it indicates that there is some other error thats preventing the analysis from proceeding, like faulty three-column format file. |
 |
|
|
Dichter
BIAC Faculty
190 Posts |
Posted - Jul 30 2007 : 6:46:34 PM
|
Thanks so much. The log file simply ends just before the "/bin/mkdir reg" command should appear. Re-running a run "manually" with the same fsf file appears to correct the problem.
Apparently, large batch jobs don't always go smoothly. Have others encountered this? |
Gabriel S. Dichter, PhD
UNC Departments of Psychiatry & Psychology
http://www.can.unc.edu/ |
|
|
|
Dichter
BIAC Faculty
190 Posts |
Posted - Jul 30 2007 : 7:30:23 PM
|
| So, I killed all jobs, deleted everything, and started the batch script over. Once again, the first lower-level analysis folder had no 'reg' folder (the line "rm -rf prefiltered_func_data*" is the last thing in the log file), but running the analysis via the GUI with the identical fsf file created the reg folder. Has anyone else encountered this?? |
Gabriel S. Dichter, PhD
UNC Departments of Psychiatry & Psychology
http://www.can.unc.edu/ |
|
|
|
syam.gadde
BIAC Staff
USA
421 Posts |
Posted - Jul 30 2007 : 7:59:04 PM
|
Not sure why identical .fsf files are doing different things...
For the failed run, the report.log file will tend to have errors listed in it. Generally you want to find the first error that crops up there. Any clues there? |
|
|
|
vinod
Average Member
USA
68 Posts |
Posted - Jul 30 2007 : 8:19:16 PM
|
If you modify the .fsf manually from outside GUI, this is highly possible. There are a lot of flags in the .fsf file which the GUI automatically updates, but you overlook when you manually change. For instance, you can go into the .fsf file and say "set fmri(analysis) 6" to run stats + post-stats only. But unless you also go and change the "set fmri(filtering_yn) 1" to 0, it will still perform pre-processing. But if you run through the GUI, the flag gets automatically updated. I learnt this the hard way :).
So, I would recommend checking all the registration based flags in your .fsf file. For example, there is a flag called reg_yn, make sure that is not a zero. SAme with other flags like reginitial_highres_yn and regstandard_yn. Another way to look for differences between what happens with the .fsf file you created and the design.fsf file which is created when running from the GUI using the "diff" command. |
|
|
|
Dichter
BIAC Faculty
190 Posts |
Posted - Jul 30 2007 : 11:20:11 PM
|
| Vinod saved the day: my template fsf file had fmri(poststats_yn) set to 1. Apparently that setting was causing problems via batch scripts but not via the GUI. Setting it to zero resolved the issue. Thanks, Vinod! |
Gabriel S. Dichter, PhD
UNC Departments of Psychiatry & Psychology
http://www.can.unc.edu/ |
|
|
|
vinod
Average Member
USA
68 Posts |
Posted - Jul 30 2007 : 11:26:51 PM
|
No worries :). Everytime we change analysis types from the standard, I guess its better to generate the template again using the GUI for the type of analysis we want to run rather than modify the previous template. It is definitely not obvious how these flags are set for these analysis. Like there in no way its obvious why the poststats_yn = 1 will not do registration but poststats_yn = 0 will do registration.
I still envy how you tried out pre-stats (motion correction and slice timing correction) in less than 10 mins. Mine takes at least an hour if not more on golgi. I knew golgi was slow but didnt think the performance difference was this drastic! |
|
|
|
Topic |
|
two antomical images for registration
