Dear Christophe, All,
> a quick comment about your previous comment
> "If you look at your whole-brain results with an FWE corrected voxel
> threshold of .05, then you already correct for multiple comparisons on
> voxel level. Thus, you don't have to look at cluster statistics at all
> (typically one uses some extent threshold like k > 10 voxels though to
> avoid "single voxel" findings)."
>
> Well, if a single voxel survives the p<.05 FWE corrected threshold, then
> it should certainly be reported as it IS significant after correction.
> The "k>10 voxels" rule does not have much ground here and actually I
> wonder how you would explain that you're discarding significant
> results... Moreover consider data acquired with 2x2x2 mm^3 voxels and
> resampled during normalisation at either 3x3x3 mm^3, 2x2x2 mm^3 or 1x1x1
> mm^2, then 10 voxels could mean very different things.
I think this is an interesting point where statistics meets reality. As
Helmut, I am also in favor of using an extent threshold even after
correcting for multiple comparisons, for the same reason Helmut
mentioned: I simply do not believe that a single voxel, however
statistically significant, is biologically plausible. Yes, it could be
the tip of the iceberg and there could be a meaningful cluster just
below the threshold (this is why I always look at lenient thresholds as
part of my results exploration), but sometimes (and more often than not)
it will simply be chance, arising from low variance, not high effect size.
I fully agree that the spatial resolution should be factored into this,
but if you consider the more typical case of originally 3x3x3mm data
being resampled to 2x2x2mm, then 10 "active voxels" in native space
(already a small cluster, with a volume of about 1/4 ccm!) is
represented by about 34 voxels in normalized space. In the puristic
view, 1 of these voxels should be accepted if significant, but this only
represents about 1/3 of a "true" voxel in native space. I have a hard
time believing that other people's fMRI data is so good to be that
trustworthy (mine sure is not :)
I am not sure I have a good answer to how to address this, or a good
rule of thumb, but I will continue to argue that it is legitimate to
factor biological plausibility into the decision on using an extent
threshold. Looking forward to hearing other people's take on this.
Cheers,
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/
____________________________________________________
|