Dear FSL community,
We are trying to set up the model in Feat for an experiment that
includes some subjects with no information for certain conditions.
We're using the custom 3-column format for the EVs. For conditions
with no information, the EVs contain "0 0 0". For those EVs with no
information, we're convolving with the default flobs optimal basis
functions. In the best case, this results in the expected warning
about the matrix being rank deficient, and that's ok.
The problem is that sometimes, we get the following error:
"F-test 26 isn't valid - each included contrast cannot be a linear
combination of the others. Child process exited abnormally", where the
number of the invalid F-test corresponds to a specific set of contrasts.
A bit more information: Each condition has 3 EV files, (1), (2), (3),
of which any combination could be "0 0 0". The F-test is performed
across the 3 contrasts from those 3 EVs. In one specific example, the
first condition has 3 "0 0 0" EV files. When I try to set just that
one up, things work fine. However, when I get to condition 4, which
also has 3 "0 0 0" EVs, meaning the EV files are absolutely identical
to condition 1, I get the above error. If I skip condition 4 and
instead include condition 5, which has 2 "0 0 0" EVs and 1 EV with
information, I get the same error. Condition 6 on the other hand,
which has 3 EVs with information, works again.
If we try a different angle of attack and include EVs that are empty,
but leave those particular contrasts blank (i.e. all zero) in the
model setup, we get the error "Problem with processing the model:
Contrast 73 is empty! Child process exited abnormally", which makes
some degree of sense, but doesn't really help very much.
I am utterly confused.
Any ideas?
Phil Opitz
[log in to unmask]
Emotion, Brain & Behavior Lab
Tufts University
|