Dear Andreas,
> would that mean that any attempt to determine %signal change relative to
> the local mean is essentially nonsense when analyzing smoothed data
> parametrically,
It just means that estimating %signal change by dividing estimated
activity by estimated mean intensity in a given voxel, based on smoothed
images, is not necessarily a good idea, especially when the smoothness
is large compared to the width/structure of the underlying tissue.
> or does GM-scaling as applied by SPM99 allow for some
> related inference?
About differences over voxels? Not really. If one uses a global value as
estimate for mean intensity, one essentially compares a parameter
contrast over voxels. The t-value, F-value or p-value is the thing one
should use for inference.
> Can the area-under-the-curve be of use in some way?
Sorry, I didn't understand this question.
Stefan
--
Stefan Kiebel
Functional Imaging Laboratory
Wellcome Dept. of Cognitive Neurology
12 Queen Square
WC1N 3BG London, UK
Tel.: +44-(0)20-7833-7478
FAX : -7813-1420
email: [log in to unmask]
|