Hi Marko,
thanks a lot for the quick response!
As I understand it, this method is more sensitive to detect larger clusters, but small clusters may not be detected (Forman et al 1995), right? Can I use the method also if I do not know if I should expect a small or big cluster?
Can I choose any voxel-level threshold, e.g. p < 0.0001 to reduce false positives already in the first step? Or are there rules how to choose the threshold?
I'm afraid I do not really understand how CorrClusTh works - I was a bit suprised that the result comes immediately without a number of iterations... Is there a link for further information or could you explain it a bit? Thank you!!!
Best wishes
Janina
-------- Original-Nachricht --------
> Datum: Tue, 19 Jul 2011 09:53:22 +0200
> Von: Marko Wilke <[log in to unmask]>
> An: [log in to unmask]
> Betreff: Re: [SPM] CorrClusTh_Cluster-size_correction
> Hi Janina,
>
> > I would like to compare fMRI data of two groups (14+14) in a 2nd
> > level analysis in spm. Is it OK to do a cluster-size correction with
> > http://www.sph.umich.edu/~nichols/JG5/CorrClusTh.m by Nichols and
> > Wilke (or other tools compatible with spm) INSTEAD of FWE or FDR?
>
> Obviously, the complicated stuff in that script was conceived and
> written by Tom :) but in any case, the idea is to generate a cluster
> size that "ensures" (at the specified level of significance) that these
> clusters did not arise by chance. As you can see, you first have to
> specify a voxel-level threshold, which is then used to derive the
> cluster size. The question as to whether you can use this approach
> instead of FWE or FDR is not an easy one to answer as these approaches
> control different things: the "regular" FWE/FDR approaches control the
> error on the voxel level, while the script determines the cluster size
> that, in combination with a (likely less-stringent) voxel-level
> threshold, results in a cluster-size corrected result.
>
> > What does the value "Expected voxels per cluster" in the spm results
> > table mean (as this is compleatly different from the value
> > calculated with the tool).
>
> This refers to the estimated smoothness of the images that is needed to
> calculate the correction for multiple comparisons on the voxel level,
> based on the Gaussian random field approach as used by FWE (see Matthew
> Brett's excellent tutorial on this one, at
> http://imaging.mrc-cbu.cam.ac.uk/imaging/PrinciplesRandomFields).
>
> Hope this helps,
> Marko
>
> --
> ____________________________________________________
> PD Dr. med. Marko Wilke
> Facharzt für Kinder- und Jugendmedizin
> Leiter, Experimentelle Pädiatrische Neurobildgebung
> Universitäts-Kinderklinik
> Abt. III (Neuropädiatrie)
>
>
> Marko Wilke, MD, PhD
> Pediatrician
> Head, Experimental Pediatric Neuroimaging
> University Children's Hospital
> Dept. III (Pediatric Neurology)
>
>
> Hoppe-Seyler-Str. 1
> D - 72076 Tübingen, Germany
> Tel. +49 7071 29-83416
> Fax +49 7071 29-5473
> [log in to unmask]
>
> http://www.medizin.uni-tuebingen.de/kinder/epn
> ____________________________________________________
--
Empfehlen Sie GMX DSL Ihren Freunden und Bekannten und wir
belohnen Sie mit bis zu 50,- Euro! https://freundschaftswerbung.gmx.de
|