Dear Karl,
thank you for your reply.
> > Search volume: S=408686mm^3 = 8715 voxels = 363.0 resels
> > Voxel size: [2.2, 2.2, 9.8]mm [1 resel = 2.56 voxels]
> >
> > I cannot understand,
> > (1) why the ratio, S / resel, is 363.0 (2nd line from the last), instead
> > of 3404 expected from the last line.
>
> The number of RESELS is now computed using the actual intracrianial
> voxels as described in Worsley KJ et al 1996, Hum Brain Mapp. 4:58-73.
> This estimate is not simly the S divided by prod(FHWM), although they
> should be nearly the same. I agree the discrepancy in your estimates
> is remarkable, and may reflect the fact that (i) you have lots of
> little holes in the data or (ii) the fact your data do not conform to
> a lattice representation of a smooth field. I would smooth the data
> such that the smoothness is at least 2 voxels.
I applied smoothing (FWHM=4x4x4mm and 4x4x12mm) to my data, and found that
S=1616.8 resels and 2076 prod(FWHM) for 4x4x4mm smoothing and
S=1485.9 resels and 1764 prod(FWHM) for 4x4x12mm smoothing, both having
smaller difference than
those of 363.0 resels and 3404 prod(FWHM) for unsmoothed data, as you suggested.
> The computation of corrected P values has been refined considerably to
> use RESEL counts in all dimensions (see Worsley KJ et al 1996, Hum
> Brain Mapp. 4:58-73.) The marked discrepancy you observe is due to the
> fact that the assumption of a good lattice approximation has been
> violated. I would repeat your analyses after smoothing the data with
> a 4mm FWHM spatial filter,
>
> If you are interested see spm_P.m and the subfunctions it calls.
I understood that E{m} is substituted by a sum of (Resel)D times
(EC densities)D for D=0-3 in spm_P.m, as Eq.3.1 of Worsley et al 1996.
Sincerely,
fujimaki
--
Norio Fujimaki, Ph.D
Brain Function Research Group
Auditory and Visual Informatics Section
Communications Research Laboratory
588-2, Iwaoka, Iwaoka-cho, Nishi-ku, Kobe, Hyougo 651-2492, Japan
Tel:+81-78-969-2261
Fax:+81-78-969-2269
e-mail: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|