Dear Zach,
if you change this option in spm_defaults you can use the non-stationarity correction in SPM12 that relies on local resel values:
defaults.stats.rft.nonstat = 1;
This correction is based on:
http://www.fil.ion.ucl.ac.uk/spm/doc/papers/HayasakaNichols_NonstCls.pdf
Best,
Christian
On Sun, 21 Feb 2016 15:33:22 -0700, Zach Christensen <[log in to unmask]> wrote:
>Dear SPM Experts,
>
>I'm currently developing some functions for R statistical software that
>utilize random field theory for MRI. I've been able to successfully compute
>results comparable to those from SPM12 on some VBM images but I wanted to
>include functionality that accounts for non-isotropy. After reading through
>some of the literature I looked at the FMRISTAT and SurfStat toolboxes to
>compare approaches for these computations but do so with some trepidation
>because neither seems to be very up to date. My current understanding is
>that SPM doesn't fully account for non-isotropy because by default it
>doesn't use the resels per voxel image but instead uses a conversion based
>on the total FWHM estimate. (I think one of the papers about this approach
>is 'Detecting sparse signals in random fields, with an application to brain
>mapping'.)
>
>However, I recently came across the paper 'Set-level threshold-free tests
>on the intrinsic volumes of SPMs' that implies these features should be a
>part of SPM12. I understand that the latter paper refers to a different
>process than that which I previously mentioned, but I can't seem to find
>full support in SPM for either method. Essentially I'm asking if there is a
>way for me to compute results accounting for non-isotropy or the related
>set-level threshold-free test using any versions of SPM. If not is their
>any software out there that currently supports these features?
>
>Thank you
>
>--
>-Zach Christensen
>
|