Dear Ben,
Unless you change the defaults, SPM implements FDR on topological
features of the underlying random field, ie cluster extent or peak
height as described in these two publications:
http://dx.doi.org/10.1016/j.neuroimage.2009.10.090
http://dx.doi.org/10.1016/j.neuroimage.2008.05.021
So with topoFDR = 1 the relevant column in the results table will
contain peak-level FDR q-values.
If you change topoFDR to be 0, this column will contain voxel-level FDR
q-values (the heading is unmodified, hence the confusion?).
Best regards,
Guillaume.
On 06/02/17 23:31, Benjamin Sinclair wrote:
> Dear All,
>
> I want to do an FDR correction at the voxel/peak level, as opposed to the cluster level.
> To do this I set defaults.stats.topoFDR = 0.
>
> However the SPM results window will display a peak-level FDR q value (under heading: peak-level: qFDR-corr) regardless of whether defaults.stats.topoFDR = 1 or defaults.stats.topoFDR = 0.
> But the peak-level FDR q value is different depending on whether defaults.stats.topoFDR = 1 or defaults.stats.topoFDR = 0.
>
> Is the peak-level FDR q value invalid if defaults.stats.topoFDR = 1?
> If so, why is it displayed? What does it represent?
>
> Cheers
> Ben
>
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|