Dear Ning Ma,
> I heard about the FDR corrected way was not very good in SPM5, so we have to
> use FWE for the correction. Since changing to SPM8, we wonder in SPM8, if
> the FDR correction way is valid and good enough to use?
FDR correction can be applied in different ways. In SPM5, FDR was
commonly used as a voxelwise correction. Chumbley and Friston (2009)
argued that it makes more sense to apply FDR at the cluster level;
i.e., in a way that takes into account that fMRI data has some spatial
structure. For that reason, in SPM8, voxelwise FDR Is not available
by default, but cluster-level correct p values calculated using FDR
are given. It is still possible to obtain the option for voxelwise
FDR by changing the values in spm_defaults (set
defaults.stats.topoFDR=0). So, I don't think it's really that
voxelwise FDR (as implemented in SPM5) was incorrect, but that there
are different ways of thinking about fMRI data, and the SPM developers
tend to favor the topological implementation of FDR, as it exists in
SPM8. There are a number of other posts on this topic (try searching
for "SPM8 FDR" or "defaults.stats.topoFDR").
All of which is a long way of answering your question: namely, FDR
correction in SPM8 is certainly a reasonable way of controlling for
multiple comparisons! :)
Reference:
Chumbley JR, Friston KJ (2009) False discovery rate revisited: FDR and
topological inference using Gaussian random fields. NeuroImage
44:62-70.
Hope this helps!
Best regards,
Jonathan
--
Dr. Jonathan Peelle
Department of Neurology
University of Pennsylvania
3 West Gates
3400 Spruce Street
Philadelphia, PA 19104
USA
http://jonathanpeelle.net/
|