> I've bee analyzing my subjects with and without normalisation. Some of the
> normalized maps, however, show activations that are not present in the
> non-normalized ones (being all other analysis steps done the same way). Is
> there any chance that the these extra activations might indeed represent
> actual activations? (Or should I assume that they're just artifacts?)
I would expect the pattern of activation to change slightly after spatial
normalisation. This may be by just enough to make p values significant, that
were not significant previously, or maybe just enough to lose significance.
Are there any surprises in the unthresholded statistic images? Don't ask me
about interpretation though.
There are many possible reasons for these differences, including: additional
smoothness of resampled data, so different smoothness estimates; different
field of view, so different estimate of "globals", leading to a different
grey matter threshold and different number of voxels analysed; expansion nd
contraction of spatially normalised images, leading to different mixture of
local intensities after smoothing.
Best regards,
-John
|