[FieldTrip] cluster-based permutation test on time-frequency data

[David Groppe]

Hi Shen-Mou,
      Cluster-based permutation tests have great power for detecting
broadly distributed effects, but this comes at the cost at being less
sensitive to less broadly distributed effects.  A good alternative to
cluster-based permutation tests is Benjamini & Hochberg's false
discovery rate (FDR) control algorithm.  Like cluster-based
permutation tests FDR control provides weak control of family-wise
error, and we've found that it has relatively good power for broadly
and narrowly distributed effects:

Groppe, D.M., Urbach, T.P., & Kutas, M., Mass univariate analysis of
event-related brain potentials/fields I: A critical tutorial review,
Psychophysiology, 2011, DOI: 10.1111/j.1469-8986.2011.01273.x
http://kutaslab.ucsd.edu/people/kutas/pdfs/2011.P.01273.pdf

I'd recommend using that FDR control procedure over the cluster-based
permutation procedure unless you are only interested in the broadly
distributed effects.
     -David


On Thu, May 10, 2012 at 12:31 AM, Shen-Mou Hsu <explena at gmail.com> wrote:
> Dear Users,
>
> I wonder if anyone could provide any suggestion regarding an issue I
> encountered. According to the article (Maris, 2007), it seems that
> cluster-based permutation test is calculated under the permutation
> distribution of the maximum cluster-level statistics. In my tome-frequency
> data, there are two big clusters. Only one cluster reaches statistical
> significance; however, if I exclude this significant cluster and then rerun
> the test, the other cluster also reaches statistical significance. It seems
> to suggest that the sensitivity is lost for the second cluster. I wondered
> if there is any approach to resolve this issue. Many thanks!
>
> Best regards,
>
> Shen-Mou Hsu
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip



-- 
David Groppe, Ph.D.
Postdoctoral Researcher
North Shore LIJ Health System
New Hyde Park, New York
http://www.cogsci.ucsd.edu/~dgroppe/