Hi Ged,
> Something is still not quite clear to me: even if this is fixed, will
> the RF/FWE thresholding generally use the nonstationary RPV values, or
> is it only the cluster p-values from spm_list that will take account
> of this?
As for the FWE-corrected threshold, probably the RPV values aren't used.
The threshold is determined by the number of RESELs for the entire
brain.
As for the cluster p-values, non-stationarity is accounted by the RPV
values somewhat, if the 'VRvp' (sic) bug is fixed. One thing to note is
that, even if the bug is fixed, SPM uses the RPV value at the maxima,
which may or may not reflect the smoothness/roughness of a given
cluster. The RPV image itself is a random field, and Keith Worsley's
stat_threshold can properly account for the variability in the RPV
values.
-Satoru
|