> with increasing number of subjects per study and more and more contrasts of
> interest spm GUI second level analysis (simple or paired t-test) takes a
> lot of time and gets highly error vulnerable when picking the con_*.img
> files by hand (e.g. by accidentally choosing a wrong contrast in one
> subject).
The 'Edit' button to select files is very useful in this situation:
If the contrast of interest has the same number for each subject (as
should be the case if you used the same batch to generate the contrasts
for individual subjects), then a simple strategy is to use a unix shell to
list the relevant con*.img files, e.g.:
ls -1 $PWD/*/stats/con_0004.img
When SPM's file selection box appears, click on the 'Edit'
button, and cut & paste the list of files.
Once the model has been estimated, you can use the batch system to
generate the contrasts of interest.
If you are looking at, say, a dozen of contrasts at the second level, this
trick is acceptable.
If you have 100 contrasts, then, yes, a batch system would be better.
Christophe Pallier
INSERM U562 SHFJ/CEA, Orsay France
PS: unfortunately, 'ls' is not a native command of Windows. I don't know
how to generate a proper list of file with "dir" (?).
|