Curious. Perhaps running through the gui is correcting some problem in your sed'd fsfs? Make copies of your fsf files and diff them against those the gui writes into your feat directories.. Hopefully there'll be a difference that will explain all.
------Original Message------
From: Stephane Jacobs
Sender: FSL - FMRIB's Software Library
To: [log in to unmask]
ReplyTo: FSL - FMRIB's Software Library
Sent: 2 Dec 2008 18:59
Subject: [FSL] batch scripting higher level analyses
Dear FSL users,
I've been trying to find previous reference to this problem in the archives
but could not find anything relevant. I'm trying to run higher level
analyses (across subjects) using batch scripts (either bash or perl), as
I've successfully done for first-level analyses. Basically, the script edits
a template design file using sed to generate an appropriate design file, and
then calls the feat command ("feat design.fsf").
The analysis starts as it should, but fails to launch the higher level
stats. It completes without generating any error message, but no copeX.feat
subdirectory is created, and no result is actually computed.
When I call the same command from a terminal window, the same thing happens.
When I actually open the FEAT GUI and load the same design file that the
script uses, the analysis runs fine until the end and generates the expected
copeX.feat subdirectories.
Is there any reason why higher level analyses across subjects require more
than calling "feat design.fsf" fom the command line or from a script? I'm
probably missing something here...
Any help will be very much appreciated!
Thanks,
Stephane
Sent using BlackBerry® from Orange
|