Hi
Yes, reporting corrp_tstat clusetrs is fine - we simply opted for GGMM as opposed to cluster FWE correction as the hypothesised areas are very small.
hth
Christian
On 21 Apr 2011, at 12:02, Athena Demertzi wrote:
> Dear FSL people,
>
> I am currently using Dual regression on group ICA data and I have a question about statistical inference. I ordered non-parametric stats (permutation test) as implemented in the DR script and I get the output. My question is: would it suffice to report the corrp_tstat (FWE correction) when I report my contrast between my two groups or do I need to fit a Gaussian/gamma mixture model as mentioned in the Filipini, PNAS paper?
>
> "These maps were thresholded using an alternative hypothesis test
> based on fitting a Gaussian/gamma mixture model to the distribution of
> voxel intensities within spatial maps and controlling for the local
> false-discovery rate at P < 0.05."
>
> What does it add? And in case I would like to use the GGMM, how can I do that in terms of software and procedure (in a user-friendly way if possible)?
>
> Thank you very much for your time and consideration.
>
> Sincere regards,
>
> Athena
>
> Coma Science Group
> Cyclotron Research Centre
> University of Liège, BELGIUM
|