Nicola Pannacciulli wrote:
> Dear SPMers,
>
> I would very much appreciate if somebody could explain, in layman's words:
>
> 1) the difference between FDR- and FWE-corrected P-values (and between the
> cluster- and voxel-level correction);
>
If you took 20 imaging data sets that were just noise
(but otherwise had the same signal characteristics of your fMRI data)
then an FWE corrected p-value of 0.05 means that only 1 SPM image in 20
will contain a false positive.
An FDR of 0.05 means that, of the A voxels that are declared
as active, then typically 0.05*A will have been falsely declared. So if you see
1000 voxels in your FDR-0.05 corrected image, then about 50 will be false positives.
> 2) which one (cluster-level, voxel-level, FDR-corrected, FWE-corrected?)we
> are supposed to take into account when reporting our results.
>
Ideally people like voxel-level inference as its more anatomically precise than
cluster-level inference. But if they can't get significant results at the voxel-level
, then the cluster-level is a useful fall-back.
As to what you *should* use, I can't say. Historically, the advice has been
to always try for voxel-level FWE 0.05. But there are many papers out these
days using FDR. As long as you adequately describe the approach you've used and
what it means, you're on safe ground.
You may wish to read papers on this topic from the SPM Online Bibliography
http://www.fil.ion.ucl.ac.uk/spm/doc/biblio/Keyword/RFT.html
Best wishes,
Will.
> Any suggestions and hints will be highly appreciated. Thank you very much.
>
> Regards,
>
> Nicola Pannacciulli
>
>
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|