Ok, I emailed the report and ls -lR. Here is a possible issue:
It's possible that since avwmerge is trying to merge together 102
masks/copes/etc. that it runs out of memory and fails to produce the merged
mask?
I tried running a small group analysis that basically merges the three
individual subject runs into a single group. This worked beautifully, so
the first level feats must be fine.
Which brings up another point:
If I created a gfeat for each subject that merged the 3 runs together, and
then ran a gfeat for every contrast (using the copes from the individual
subject gfeat), would this be statistically valid?
I understand the benefit of combining all of the runs (102) in one giant
2nd level analysis would be a better estimate of variance, so would this
benefit be negated by performing individual subject gfeats (to combine
runs) followed by a true group gfeat?
Thanks,
Tyler
|