Hi,
I was having the same problem and I realized that for my case there was
something funny in SPM when reviewing the 2nd level design SPM.mat file. I
am running a single-factor, 6-level design with 16 subjects per cell.
However, when reviewing the design in the SPM gui (Design->Explore->Files
and Factors), I realized that the level (1 through 6) and subject number (1
through 16) were assigned wrongly to the contrast file paths, as if they had
been shuffled around (as if bootstrapping). In my case, this was solved by
changing line 39 in file spm_set_factorial_design.m
from
P=[P;job.des.fd.icell(cell_num).scans];
to
P=[P;job.des.fd.icell(cell_num).scans'];
so the correspondence now is correct. Getting this right allows me now to
run the 2nd level estimation without any problem, yielding reasonable results.
This might be a general bug in SPM5, in which case similar changes would
probably be needed in two and three-way designs.
Cheers,
Albert
|