 |
 |
Hi David, Yes, pre-threshold masking with voxel-level thresholding should be similar to small-volume correction as in SPM2/5/8 (not sure if things have changed for SPM12; others in the list could probably tell). But there will be differences in results, not just because of the overall modelling of the temporal correlation and estimation of smoothness, but also because when it comes to RFT, SPM uses the Worsley et al (1996) equations, that use results for each particular statistic (t, F, z), whereas FSL converts t and F to z, and uses the equations for z. Also, SPM takes into account the lower dimensional terms of the expression for the Euler characteristic; these are much smaller than the one for the 3rd dimension and are usually negligible, whereas FSL uses the 3rd and (much) larger one. All the best, Anderson On 9 September 2015 at 15:35, David V. Smith <[log in to unmask]> wrote: > Hi -- just following up on this message. Would "pre-threshold masking" > combined with "voxel level thresholding" be similar to the small correction > in SPM? My goal is to correct for the number of comparisons within a > specific mask, but I don't want to do a Bonferroni since the > voxels/comparisons within that mask are not all independent. > > Thanks! > David > > > > On Sep 3, 2015, at 2:01 PM, David V. Smith <[log in to unmask]> > wrote: > > > > Hi, > > > > I might be off base, but is it correct to assume that the "pre-threshold > masking" + "voxel level thresholding" options in FEAT is equivalent to the > small volume correction in SPM? > > > > Also, how are the "number_of_resels" computed for this approach? I'm not > 100% sure where it is pulling the value it used in the -g option for ptoz > in the feat logs: > > > > /usr/local/fsl/bin/fslmaths thresh_zstat4 -mas > /path/to/the/prethreshold/mask.nii.gz thresh_zstat4 > > > > echo 333 > thresh_zstat4.vol > > zstat4: DLH=0.81907 VOLUME=333 RESELS=5.64651 > > > > /usr/local/fsl/bin/ptoz 0.05 -g 58 > > 3.887894 > > > > I thought the ptoz command would use the number of resels reported in > the "stats/smoothness" file. Or is that the size of each resel, making the > number_of_resels = VOLUME/RESELS ?? > > > > Thanks! > > > > David >