On Sun, 22 Mar 2009 10:22:47 +0100, < Marina> < Lopez-Sola>
<[log in to unmask]> wrote:
> Thanks very much for your help. Any more info to understand how the non-
> sphericity correction is performed and how the different variances among
> the three conditions can affect the final comparison among them will be
> super appreciatted. Or any source of info on that issue.
I looked at one example that I did in my own work.
The design matrix that doesn't reflect the correction (at least in SPM5) is in
SPM.xX.X
The design matrix _with_ the correction, and which (at least in my example)
has the same values as those displayed by clicking on the design matrix
schematic produced when you "review" the design, is in
SPM.xX.xKXs.X
_What_ the nonsphericity correction is, and _how_ it works, is a long and
deep topic. You can get started by reading the comments in the SPM code.
The file spm_spm.m has a brief discussion.
For a list of papers, you can look at the online bibliography at the SPM site.
The link for those with keyword "nonsphericity" is
http://www.fil.ion.ucl.ac.uk/spm/doc/biblio/Keyword/NONSPHERICITY.html
Some of the stuff there, however, is concerned with subject-level fMRI time
series, which isn't at issue here (though some of the theory overlaps, I
assume).
Warning: this stuff is very technical.
|