Dear Philipp,
> for VBM-like processed DTI-data, could the SnPM toolbox be used for
> non-parametric analysis?
> (as non-Gaussian distribution of residuals tends to be present in the
data
> even after smoothing)?
Are you interested in comparing fractional anisotropy or mean
diffusivity images, one image per subject? Then yes, it is possible. The
permutation test in SnPM does not require the distribution to be
Gaussian.
> If yes, what would be the settiings for
> (a) a two-group comparison and
> (b) for a simple correlation over the scans of one group using one ,
> ideally two covariates (one cov. of interest, one nuisance variable)?
>
For (a), you should use "MultiGroup: 2 Groups, 1 Scan Per Subject'. For
(b), you can only use one covariate at a time, so you cannot enter
nuisance covariate. I am currently working on a permutation test with
one continuous covariate of interest and multiple nuisance covariates.
Hope this helps!
-Satoru
|