Hello everybody,
We ran repeated measures ANOVA (Full Monty under PET) in
SPM2 on fMRI data consisting of two groups of subjects with 3 scans for 3
conditions (1 per each condition). Other options that I picked were “no
global normalization”, “scaling of overall grand mean to 50”
and “mean voxel value” global calculation.
The results looked a little suspicious, so we decided to
look deeper into it. We compared voxel values (at the most significant voxel)
plotted by SPM with values manually extracted directly from image files from
the same coordinate. The plots don’t even look similar, the values seem
to be scaled, but even the pattern is different. So my question is what kind of
adjustments/calculations are done on voxel values before t-scores are
calculated? We did repeated measures ANOVA in SAS on those values, and the
t-score is nowhere near what SPM displays.
To make sure that my extracted values are correct, I did a
similar thing for a two-sample t-test analysis, and the plots of my extracted
values and that of SPM look almost identical (with SPM plot shifted to center
it at zero).
Thank you beforehand.
Regards,
Renat.