Hi Julia,
You really should not go below z=2.3 as a cluster forming threshold, and
in fact Tom Nichols recently suggested using z=3.09 as a lower limit for
the cluster forming threshold in order to satisfy the approximations
used as part of the underlying Gaussian Random Field Theory.
See this post for details and references:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?
A2=ind1105&L=FSL&D=0&1=FSL&9=A&J=on&d=No+Match%3BMatch%
3BMatches&z=4&P=92435
cheers,
-MH
On Wed, 2011-06-29 at 11:49 -0400, Julia Foecker wrote:
> Dear Mark,
> thanks a lot, that was really helpful!
> Actually, I am analyzing one subject and Looking at the data.
> When I use the z-threshold in the poststats, there is this default value of 2.3.
>
> Actually at this threshold I do not find much activation...is there something
> like a "lower boundary" which one has to keep, e.g. 1.5 or sth. like that?
>
> Thanks a lot,
> Julia
>
>
>
>
> On Wed, 29 Jun 2011 13:51:52 +0100 Mark Jenkinson <[log in to unmask]> wrote
>
> > Dear Julia,
> >
> > I assume you are referring to the Misc tab in the FEAT GUI
> > and in particular to the "Z Threshold" within the "Design efficiency"
> > part. This is not a threshold that is used to calculate your
> > activation statistics in any way, so do not be concerned. This
> > simply refers to the efficiency calculation (like a power calculation)
> > that is presented with the design matrix, showing the % BOLD signal
> > change required for each contrast. I would leave this at its default
> > setting as it gives you a reasonable estimate for efficiency (relating
> > more to the sensitivity to real effects rather than the null-hypothesis
> > false positive rate).
> >
> > The actual thresholds for your statistics are found in the Post-stats
> > tab and are controlled via the corrected p-values (also with an
> > arbitrary z-threshold if cluster-based correction is used). It is only
> > the threshold in the Post-stats tab which will be used on your data
> > to determine statistically significant activation.
> >
> > All the best,
> > Mark
> >
> >
> > On 29 Jun 2011, at 13:32, Julia Foecker wrote:
> >
> > > Hi,
> > >
> > > I have a question concerning the Misc option in FSL.
> > > The z-threshold was preset to 5 (or 5.8).
> > >
> > > How do I know which is the best threshold for my data? (I think the 5.8 is
> > > much
> > > too high...).
> > > And are there any values which should not be used because they are too low?
> > >
> > > Thanks a lot,
> > > Julia
> > >
|