[FieldTrip] cluster based permutation test - cfg.clusteralpha threshold

[Aaron Schurger]

Dear Balint,
It was difficult to parse your question: "it’s a valid way to choose a
more permissive threshold (cfg.clusteralpha) at the level of the
clusters identifying t-tests?" If you are asking whether or not it is
valid to use a lower point-wise threshold, then it is technically
valid, but if, for example, you try every possible point-wise
threshold from 0.01 to 0.20 in steps of 0.01, and you manage to find
one that gives you a significant result - well that should be avoided.
Otherwise it is OK. You might also try this:
http://www.biomedcentral.com/1471-2202/14/122 with code available
here: https://bitbucket.org/emsf/emsf_matlab

Instead of trying to find clusters of points that covary with an
experimental effect, you try to find a temporo-spatial weighting that
maximizes the experimental effect, and then look at the spatial
patterns (clusters) that give you the effect.

Best,
Aaron

On Thu, Apr 24, 2014 at 4:23 PM, Balint File <file.balint at ttk.mta.hu> wrote:
> Dear Fieldtrip experts,
>
> Recently I started to apply the spatial dimension (electrode neighborhood)
> of the cluster based permutation test. My question is: it’s a valid way to
> choose a more permissive threshold (cfg.clusteralpha) at the level of the
> clusters identifying t-tests? According to the publication of Maris et al.,
> 2007 („this threshold does not affect the FA rate of the statistical test.
> However, this threshold does affect the sensitivity of the test. For
> example, weak but long-lasting effects are not detected when the threshold
> is large”) it seems to me, that this threshold depends on the nature of the
> data. For lower cfg.clusteralpha  threshold (<0.15) I get more extensive
> (and more explainable)  results and the clusters still show high
> significance (<<0.025) at the permutation test.
>
> Thanks in advance for your help.
>
> Best regards,
>
> Bálint File
>
>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip



-- 
Aaron Schurger, PhD
Senior researcher
Laboratory of Cognitive Neuroscience
Brain-Mind Institute, Department of Life Sciences
École Polytechnique Fédérale de Lausanne
Station 19, AI 2101
1015 Lausanne, Switzerland
+41 21 693 1771
aaron.schurger at epfl.ch
http://lnco.epfl.ch/