Dear Barbara,

You could have a look at the talk "Random Field Theory & Alternatives", by Tom Nichols, available here.

Regarding your question here under "Is it a problem to go by e.g. p<0.01; 3500 voxel cluster threshold and p<0.05 FWE cluster correction?", this is not right fr 2 reasons:

- you should not be using a "cluster forming threshold" higher than .001, as this will lead to not-so-well-controlled false positive rate on the cluster extent (see original papers on Gaussian field theory or the recent Eklund et al. paper)

- setting a cluster size threshold of 3500 is rather arbitrary and doesn't reflect a useful statistical threshold. The number of voxels needed to reach p<.05 FWE depends on SPM at hand. See this bit of code for example.

Best,
Chris

-

De: "Barbara Kreilkamp" <[log in to unmask]>
À: [log in to unmask]
Envoyé: Lundi 22 Mai 2017 17:24:15
Objet: Re: [SPM] Fwd: [SPM] extract large number of p-values from clusters

Hi Rouhi and SPM experts,

Thanks. Do you know about the 'topological interference' and can explain it to me please?

Thanks a lot,
best wishes,
barbara

[log in to unmask]">

Dear Barbara,

I have written a small script to do that. You may check if it works for you or not.

Best

Rouhi

Best Regards

Rouhollah Abdollahi

Institut für Neurowissenschaften und Medizin (INM-3)
Forschungszentrum Jülich
Leo-Brandt-Straße
52425 Jülich, Germany

Tel: +49 2461 61-4686

On Fri, May 19, 2017 at 4:35 PM, Barbara Kreilkamp <[log in to unmask]> wrote:

Thank you Guillaume, I will try this.

Meanwhile would you mind explaining the 'topological interference' thing for me? Is it a problem to go by e.g. p<0.01; 3500 voxel cluster threshold and p<0.05 FWE cluster correction?

Cheers,

Barbara

On 19/05/2017 16:08, Guillaume Flandin wrote:

Dear Barbara,

Why not using the higher level module from the batch interface:
   SPM > Stats > Results Report
There you can save the thresholded map as a NIfTI image, as well as the
results table in CSV or Excel format. This can take place without any
interaction and without graphical display if you set the defaults
accordingly:
   spm_get_defaults('cmdline',true);
After the execution of the job, the TabDat variable is also available in
the MATLAB workspace, allowing you to access the same information than
the one you got so far, e.g. TabDat.dat{1,3}.
Note that for topological inference in SPM, the cluster forming
threshold should not be set lower than p<0.001 uncorrected.

Best regards,
Guillaume.


On 19/05/17 15:00, Barbara Kreilkamp wrote:

Dear Rouhi,

Thank you very much, this seems to work quite well. However, I'd like to
avoid the interactive thing, because I have quite a few clusters to
extract, all at p<0.01 and a specific voxel cluster threshold.

Is there an SPM function that does not do this interactively or do I
have to reprogram myself?

This is what I've got so far:

[hReg,xSPM,SPM]=spm_results_ui % just to get hReg and xSPM for spm_list
- but too interactive

spm_list('ListCluster',xSPM,hReg,[1,8,'ALLORNOTHING'])
ans.dat{1,3}


Thanks,
Barbara


On 19/05/2017 11:33, Rouhollah Abdollahi wrote:

Dear Barbara,

Try spm_list() function.

Best regards
Rouhi

Best Regards
Rouhollah Abdollahi

*Institut für Neurowissenschaften und Medizin (INM-3)*
Forschungszentrum Jülich
Leo-Brandt-Straße
52425 Jülich, Germany

Tel: +49 2461 61-4686




On Fri, May 19, 2017 at 10:51 AM, Barbara Kreilkamp
<[log in to unmask]
<mailto:[log in to unmask]>> wrote:

     Dear all,

     Any hint on this is greatly appreciated.

     Thanks a lot,

     Best wishes,

     Barbara



     -------- Forwarded Message --------
     Subject:   [SPM] extract large number of p-values from clusters
     Date:      Tue, 16 May 2017 14:52:40 +0200
     From:      Barbara Kreilkamp <[log in to unmask]>
     <mailto:[log in to unmask]>
     To:        [log in to unmask] <mailto:[log in to unmask]>



     Dear SPMers,

     Is there a way to save multiple cluster p-values of different
     t-tests without continuously opening the results tab with SPM.mat?

     E.g. is there a procedure in SPM with which I can load the SPM
     file in MATLAB to extract p-values of clusters with a certain size?

     Thank you very much,

     Best wishes,

     Barbara