I have a group of 9 subjects that each underwent one trial of a conjunctive
visual search task. After running a first level analysis on each subject, I ran a
mixed effects: flame 1 and a mixed effects: flame 1+2 (both without
automatic outlier de-weighting). In addition I used fslmaths to first add all the
hr/thresh_zstat images together and then divided by 9 to get a group mean
image. I used fslmaths as a check against the results from the mixed effects.
I know that mixed effects is a statistical result whereas the fslmaths group
mean is not. When I look at these two images in fslview, both runs of the
mixed effects show greater activation than the fslmaths group mean. Why is
this? Shouldn't the mixed effects be less than the group mean due to the
statistical analysis? In addition I thought that the flame 1+2 would show less
or more accurate activation due to the low number of subjects, but it showed
more activation and even added a significant region that was not in flame 1.
Thanks,
Eric
|