Michael,
> Is there any way in SPM (or some ancillary function) to produce a
> corrected p-value based on the pairing of height threshold and spatial
> extent threshold, e.g. having p=0.001; k=50 voxels --> p=0.05 corrected,
> based on a Monte Carlo simulation.
Jean-Baptist Poline et al (NI 5:83-96, 1997) published a Random Field
method to do this, and it was implimented in SPM96. However, since it
only worked for Gaussian images (not t images) it was dropped from SPM99.
Our SnPM toolbox, with the 'Combining' option for results, does this for
basic models. Our method doesn't depend on Gaussianity, and allows the
use of other combining functions (e.g. cluster mass, as used by Bullmore
et al, TMI 18:32-42, 1999). See Hayasaka & Nichols (NI 23:54-63, 2004)
for more details.
|