 |
 |
Many thx Markus for your response !
(I would be very interested in any article you published using that procedure)
I agree that the original time-series should not be affected by the choice of the contrast but indeed they are.
Actually, the Y values extracted via the SPM VOI button seem to be affected by the chosen contrast (even if they are not adjusted)
i.e. Y values for a given voxel are not the same for a t-test A>B or an F-test like "effects of interest"...
Actually, this is why I feel uncomfortable...
Any idea of what might causes these discrepancies ? (any mistake of mine ?)
Best,
Cédric
> Message du 24/05/07 16:38
> De : "Markus Staudinger" <[log in to unmask]>
> A : "Cédric Lemogne" <[log in to unmask]>
> Copie à :
> Objet : Re: [SPM] PPI following a RFX analysis
>
> Cédric Lemogne schrieb:
> >
> > Dear all,
> >
> > In a RFX multisubjects analysis, we got an significant activation in a
> > region by contrasting two condition (say A>B).
> >
> > We now want to perform a PPI analysis to test how functionnal
> > connectivity of this region (ou ROI) differs during A vs B.
> >
> > It is quite straightforward for subjects who exhibited significant
> > voxels at the same region for A>B at the first level : We extract the
> > VOI time-series, SPM5 gives us the adequate regressors to perform a
> > PPI analysis and then we get a con.mat file to export for a subsequent
> > 2nd level analysis.
> >
> > However, there are at least some subjects in which the activation peak
> > for A>B is either very different or even absent. In this case, SPM5
> > does not allow us to extract a VOI time-series at a non-significant
> > voxel...
> >
> > We can of course choose a threshold of 1 to make all the voxels
> > significant and extract the ROI, but I am not very comfortable with that !
> >
> Hi Cedric!
> I don't understand exactly why you feel uncomfortable. Original time
> series don't get affected by choosing threshholds.
> You could go even further and evaluate an arbitrary contrast like
> effects of interest!
> In my analyses, admittedly, I excluded subjects from further PPI that
> didn't show the categorial effect on a liberal level of .05 uncorrected
> within a region specific sphere (say, 4mm for nucleus accumbens).
> Cheers
> Markus
> >
> > Any idea of how this issue could be dealt with ?
> >
> > Thank you in advance !
> >
> > Sincerely yours,
> >
> > Cédric
> >
>
>
> --
> Dipl.-Psych. Markus Staudinger
> Social Cognitive Affective Neuroscience Group
> Department of Psychiatry
> Division of Medical Psychology
> University of Bonn
> Sigmund-Freud-Strasse 25
> D- 53105 Bonn
> Germany
>
> Tel: +49 (0)228 287 19709
> Fax: +49 (0)228 287 16097
> E-mail: [log in to unmask]
>
>
>