Print

---

Dear Stefan,


Thanks for replying to my question.


On Tue, 27 Mar 2001 10:33:32 +0100
 Stefan Kiebel <[log in to unmask]> wrote:
> Dear Joe,
>
> > We are trying to understand the flow of information in
> > creating the SPM(t) values.
> > In several locations in our SPM(t) maps there were
> nonzero
> > values which in the corresponding con_.img and
> ResMS.img
> > files had NaN values,which was unexpected.
> > Are these SPM(t) values linked to the individual images
> that
> > were used in the group analysis in some way other than
> the
> > con_.img and ResMs.img files ?
>
> The SPM{t} of a contrast is comnputed by the same
> function as is the
> con_*.img. Basically, this function looks at each voxel
> that entered the
> parameter estimation and computes the contrast and the
> SPM{t}. The
> mask.img of the analysis indicates, for which voxels
> parameters were
> estimated. This means, if a voxel in the mask.img is
> zero, the con_*.img
> should have a NaN-value there and the SPM{t} a zero.


This is what I thought as well.


> This
> could be
> checked with the 'check reg' facility.


When the images are examined side by side, it is clear that
there are voxels with non-zero values in the SPM{t} that
have NaN values in the con.img.

 Could you also
> check, whether
> your con_.img and SPM{t} were generated by the same
> process? One
> indicator would be the write time of the con*.img and
> SPM{t}.


Indeed, they have the same write times.


 Or can you
> repeat the process and check, whether you observe this
> strange behaviour
> again?
>

I have deleted both the SPM*.img and con*.img files
and re-performed the analysis; the same problem
exists.

It would seem that at some point, some additional masking
is performed on the con.img files. (Also, the ResMs.img
file resembles the con.img file in that it shares the
same pattern of NaNs).

Thanks,
Joe













>
> Stefan
> --
> Stefan Kiebel
> Functional Imaging Laboratory
> Wellcome Dept. of Cognitive Neurology
> 12 Queen Square
> WC1N 3BG London, UK
> Tel.: +44-(0)20-7833-7478
> FAX :          -7813-1420
> email: [log in to unmask]