Dear SPMers,
I'm trying to understand how the random field theory (RFT) approach, to
correct for multiple testing, works in SPM. I would be very happy if
anyone could answer my questions.
1) As I've understood it, the corrected threshold from RFT only
considers the number of tests (voxels) and the smoothness of the data,
but not the test values themselves, is this correct?
If I change the regressors in the design matrix and the maximum t-test
value goes up to 14 from 10, the threshold is still the same?
2) In the literature I've seen the term "Lebesgue measure of the search
volume", is this equal to the number of brain voxels that are tested for
activity?
3) If I have calculated a volume of t-test values, should I transform
these to z-scores, calculate a corrected z-threshold and then transform
this to a t-threshold? Or should I calculate a t-threshold directly from
the t-values?
4) What is a typical level of smoothness (in voxels or mm) for fMRI data
that not has been smoothed?
5) In practice, how much smoothing must be applied for the RFT approach
to be valid? My voxels are 3.75 mm isotropic.
6) How much smoothing is normally applied to (single subject) fMRI data?
Best regards
Anders Eklund
|