Will,
I have not seen a reply to your post, so let me give you my 2 cents:
> My question is how does one come up with a rationale for setting one's
> extent cluster thresholds in SPM (e.g., using K=10 vs K=5 or even K=0).
There are of course ways to look at the significance of clusters per se,
using their extent. In other words, you can (for example, by clicking
"whole brain" in the interactive window) let SPM determine which cluster
is deemed significant based on extent. This will show you FWE-corrected
clusters and potentially FDR-corrected clusters (if you set
defaults.stats.topoFDR to 1). Note that, if you also enabled
non-stationarity correction (defaults.stats.rft.nonstat), the smallest
significant cluster may be smaller than a not-significant cluster, which
may make it somewhat tedious to find the "correct" threshold. This
approach is usually taken when not correcting for multiple comparisons
at the voxel level, as (if you do that) there is no need to additionally
correct at the cluster level.
However, this may not be what you are interested in, so perhaps you
rather want to know how, to put it bluntly, to justify getting rid of
the scattered junk/noise in the results :) My approach to this is
biological, rather than statistical: if I do fMRI and smooth the data
and then end up with a "cluster" of k = 3, this corresponds to a really,
really small piece of brain (with the common 2x2x2mm resolution in MNI
space, this is 0.008ccm of neural tissue). I take the liberty of not
believing this is credible information, although it is not trivial at
all to put a hard value to this. I usually look at the results without
an extent threshold and also look at the raw t-map (as of course, the
k=3 could be the tip of the iceberg), but usually will not end up
interpreting very small clusters. But again, this is my own biological,
not a statistical justification.
> Furthermore, does one's FWHM settings have anything to do with setting
> this K threshold?
Yes - with more smoothing, clusters will be larger, and this should be
taken into account.
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/
____________________________________________________
|