Dear Ladan,
quick question reg your processing,
is the number of voxel the same before and after normalization,
maybe you should try to reslice your data to approximate thetotal number
of voxel that you have after normalization ?!
Just a half-blind shot...
Peter
On Wed, 25 Sep 2002, Ladan Shams wrote:
> Dear Michael, thanks for your reply.
>
> > Spatially normalizing your data will effectively perform some degree of
> > motion correction which could significantly affect the statistics. Did you
> > perform any motion correction for either analysis?
>
> Yes, I compared the realigned+corregistered(+smoothing) with
> realigned+corregistered+normalized(+smoothing), and I got the same
> difference in significance (p=.05 vs p=.001). In fact, I have tried
> all combinations of various stages of preprocessing (slice
> timing--it's event-related design, realignment, corregistration, and
> smoothing) with and without normalization, and the results is the
> same. Regardless of the preprocessing steps, the activation results
> from normalized data is much more significant.
>
> Ladan
>
> > -----Original Message-----
> > From: Ladan Shams [mailto:[log in to unmask]]
> > Sent: Tuesday, September 24, 2002 6:12 PM
> > To: [log in to unmask]
> > Subject: Re: fMRI normalization and statistics
> >
> >
> > > the resampling that occurs during normalization does impose a small
> > > amount of spatial smoothing on the data, but I wouldn't expect it to be
> > > such that it would change an activation from .05 to .001.
> >
> > Yes, exactly. I did consider that it may be due to smoothing, and
> > applied smoothing to the non-normalized images, but it did not
> > help (i.e., it did not improve the significance).
> >
> > > Are you
> > > talking about corrected or uncorrected p-values?
> >
> > uncorrected p-values.
> >
> > Thanks for your reply,
> > Ladan
> >
> >
> > >
> > > On Tuesday, September 24, 2002, at 02:37 PM, Ladan Shams wrote:
> > >
> > > >
> > > > How can normalization affect the statistics? My understanding is that
> > > > normalization only does an affine transformation (and in my case, one
> > > > non-linear iteration). In other words, it only moves around the
> > > > voxels, and it does not change the value of the voxels. As such, it
> > > > should only affect the location of the activation blobs, and not
> > > > their significance. Isn't this true?
> > > >
> > >
> >
>
"Mitfrauen im Sinne der Satzung sind auch maennliche Parteimitglieder"
-----------------------------------------
Peter Erhard, Ph.D.
Neurologische Klinik und Poliklinik
Klinikum rechts der Isar
Technische Universitaet Muenchen
Moehlstrasse 28
81675 Muenchen
Germany
Phone: 49-89-4140 4686
Mobil: (49) 179-2308931
Fax : 49-89-4140 4867
[log in to unmask]
--- ACHTUNG NEU --- ATTENTION NEW ---
Theresa, Laura, Marie & Peter Erhard
Am Weiher 39
85435 Erding (Langengeisling)
Germany
Tel: 49-8122-558 600
|