Hi again,
> > If I am to understand the discussion, applying SVC to VBM is
> > (relatively) valid if an explicit assumption is made about
> > stationariness in relation to the smoothness of the regions undergoing
> > SVC relative to the average smoothness of the image (grey matter
> > partitions smoothed with 8-12mm kernel).
>
>I know this is terribly obvious, but the assumption of stationarity
>doesn't make the image any more stationary, and so does not make the
>analysis any more valid.
>
>I think what Satoru is saying is that there can still be significant
>non-stationarity in VBM analyses even at 12mm smoothing, and for this
>reason, the cluster statistic will be unreliable in SPM (but not
>fmristat). On the other hand the height correction will be fairly
>accurate - although you should correct for the smoothness within the
>ROI itself.
>
>Is that fair Satoru?
Thank you very much for clarifying :-)
If you are doing an SVC, you might want to pay attention to the smoothness
in your ROI. For example, with a spherical ROI (radius=10 voxels) and
DF=40, p=0.05 corrected height thresholds (by SPM2) are
4.3148 (FWHM=5 voxels)
3.5454 (FWHM=10 voxels)
3.1192 (FWHM=15 voxels)
2.8438 (FWHM=20 voxels)
... and so on.
-Satoru
Satoru Hayasaka ==============================================
Post-Doctoral Fellow, MR Unit, UCSF / VA Medical Center
Email: shayasak_at_itsa_dot_ucsf_dot_edu Phone:(415) 221-4810 x4237
Homepage: http://www.sph.umich.edu/~hayasaka
==============================================================
|