Hi,
melodic_IC contains the unthresholded Z-stat images. These are
generated by (i) running ICA to find the component maps and dividing
these by (ii) the residual error from the initial PCA. The importance
for this is outlined in the technical report in case you are
interested. Therefore these maps are Z (actually t-) stats in the
classical GLM sense (not to be confused with the Z- transforms used in
e.g. the 1998 HBM paper by McKeown et al. and various other
publications).
hope this helps
Christian
On 30 Jan 2008, at 12:01, Giovanni Giulietti wrote:
> Hi,
>
> I used MELODIC ICA on several fmri data sets.
> I was wondering the relevance of performing basic mathematical
> operations
> such as adding and meaning on these raw-z transformed IC maps
> (melodic_IC.nii.gz). In other words, is it correct to consider them
> as z-
> maps with the knowledge that they better fit the GGM (Gamma Gaussian
> Mixture) model?
> Thanks for your precision.
> Giovanni.
|