Dear Sherif,
> For those interested, here is a matlab program (written by Keith Worsley himself !!!), that calculates acceptable
> t thresholds as a function of: SEARCH VOLUME, FWHM,.....
>
> It's based on the Worsley et al. 1996 paper: "A unified...."
Thanks very much for posting this. I thought it might be useful to
clarify the position with small volume corrections, using:
spm99b
spm96
my vol_corr software
tstat_threshold - the Keith Worsley function in your email.
Firstly, both vol_corr and tstat work in spm 96 and spm 99b.
They both also provide corrected t statistic thresholds, for a given
alpha. tstat, unlike previous versions of vol_corr, gives a corrected
t thresholds for a sphere; I've just added the sphere calculation to
vol_corr, so they now both do this, and give the same answer (as you might
hope).
tstat, unlike vol_corr or spm99b, gives the more liberal of the relevant
random field correction and the Bonferroni correction. I think that KW has
added this refinement because, for fields of data with little smoothness, such
as high resolution unsmoothed fMRI datasets, the random field correction may be
more conservative than the Bonferroni correction.
However, for most SPM analyses, a reasonable
degree of smoothing has been used, and the random field correction is likely
to be more liberal (and more appropriate) than the Bonferroni correction.
It's worth noting that tstat gives a corrected _t_ threshold, so, for SPM96
analyses, you may want to convert this to the equivalent Z score:
spm_t2z(stat_threshold(...
The tstat code is much more compact than vol_corr.
vol_corr does some stuff that tstat does not however:
A rather hand-holdy user interface
Corrections for boxes, and volumes of interest defined by images
F, Z, SPM(Z) field corrected thresholds
Corrected p values for given F/t/Z
(this is all stuff covered by KWs 1996 paper)
You could modify tstat to do these things, if you wished.
The differences between vol_corr and spm99b are covered in the web page:
http://www.mrc-cbu.cam.ac.uk/Imaging/External/vol_corr.html
Hope that's of some use,
Cheers,
Matthew
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|