[VoxBo] permutation tool question

[Lila Chrysikou]

Thanks, Dan.
That's very helpful.
Best,
~Lila

On 10/22/07, Daniel Y Kimberg <kimberg at mail.med.upenn.edu> wrote:
>
> Lila Chrysikou wrote:
> > I was wondering if there is a specific number of subjects and/or data
> points
> > required for VoxBo to run a sufficient number of permutations.
>
> For a typical group analysis, you'd be doing a sign flipping
> permutation (not really a permutation, you flip (or not) the signs of
> your first level t-statistics, but we can call them permutations).
> There are only 2^n possible patterns of sign flipping, which for 6
> subjects gives you 64 possible permutations.  For 10 subjects, you can
> have 1024 permutations.  When you don't have enough permutations, the
> error associated with your threshold goes up.  The following formula
> (clipped from a message from Tom Nichols on the SPM mailing list a
> long time ago) gives the standard error on your p value, where N is
> the number of permutations you run.  It's supposedly a conservative
> when N gets large, but important to keep in mind.
>
>         SEp = sqrt(p*(1-p)/N)
>
> > Is there a way to fix this problem or the data are not enough for the
> > process to work with accuracy? Or are the 32/64 permutations sufficient
> to
> > obtain a usable critical t?
>
> That generally won't be enough.  So far as I know, the only remedy
> would be using RFT-based parametric statistics.  Which shouldn't be
> too bad.
>
> dan
>
>


-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Evangelia G. Chrysikou, Ph.D.
Post Doctoral Research Fellow
Thompson-Schill Lab

Center for Cognitive Neuroscience
Department of Psychology, University of Pennsylvania
3720 Walnut Street, Solomon Lab Room B51,
Philadelphia, PA 19104-6241

Office address: 3810 Walnut St.,Room 307, Philadelphia, PA 19104
215. 573.6726  (phone)
215. 898.1982  (fax)
E-mail: evangelg at psych.upenn.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.voxbo.org/pipermail/voxbo-general/attachments/20071022/5e8cfb1a/attachment-0001.html