Hello,
I am coming across a strange problem when I run fixed
effects analyses on various groups in my study, and I was hoping somebody could
tell me what is going on. This is an fMRI study being processed with
spm2.
First, an overview: Twelve subjects were run some time ago
on a particular paradigm (OLD). Recently, we ran twelve new subjects on
the same paradigm, with a small difference (NEW). I had run separate
fixed effects analyses on both groups of twelve, and was interested in running
a fixed effects for all 24 together to compare the two groups.
Now the problem: When the n=24 comparison did not look like
I expected it to, I decided to look at the two groups separately in the n=24
analysis. When I looked at the fixed effects for the 12 (old) group, the
n=24 spm looked just like the n=12 spm for that group. However, the spm
for the 12 (new) group in the n=24 analysis looked very different than the spm
for the n=12 for that group. I checked to make sure that the problem was
not simply a matter of changing the threshold, but even at a very low
threshold, a significant area from the n=12 (new) map was not visible on the
n=24 new-only spm.
Now the question(s): Is there, or should there be, a
difference between a fixed effects analysis when n=24 and you are looking at
only 12 subjects and when n=12 for those same subjects? If there should
not be a difference, which I think is true, why might I be getting different
results (I ran the analysis twice to make sure I accounted for every
scan)? If this problem is actually what is supposed to be happening,
which is the more valid method of analyzing my data? Should I do a
subjective comparison of the two n=12 analyses, or should I use the less
informative n=24 set?
Now the favor: Could whoever replies to this also send a
copy to my personal email account, since my server automatically screens out
SPM mail?
Thanks for all of your help.
Eric Murphy, M.A.
Research Assistant
Dept. of Psychiatry & Behavioral Neuroscience
(313) 577-1838