Dear Donald and Satoru,
cluster-extent FDR, as described in the Chumbley & Friston paper, is
already available in SPM8b.
It is currently disabled by default but you can easily switch to
topological FDR by changing spm_defaults.m so that:
defaults.stats.topoFDR = 1;
Satoru Hayasaka wrote:
> Actually my understanding is that FDR p-values can be calculated at
> the cluster-level as well. There was a recent paper in Neuroimage
> describing that:
> Chumbley JR, Friston KJ. False discovery rate revisited: FDR and
> topological inference using Gaussian random fields. Neuroimage. 2009
> Jan 1;44(1):62-70. Epub 2008 May 23.
> The paper is from the FIL, so perhaps the method described in the
> paper can be implemented in SPM.
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping) on behalf of MCLAREN, Donald
> Sent: Wed 3/18/2009 9:30 PM
> To: [log in to unmask]
> Subject: [SPM] FDR combined with cluster extent
> My understanding of FDR is that it corrects data at the voxel-level. Is it
> possible to relax the voxel-wise FDR-rate (e.g. to .1 or .2) and set a
> minimum cluster size such that the cluster inference is p<.05? If so, how is
> the p-value of the cluster computed?
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG