cluster-based analysis

[Emanuel van den Broeke]

Dear Michael and others,
I am thinking about an alternative cluster based statistic, but do not know if 
this is also valid.

The alternative method goes as follows:
1.	Calculate t-statistics of two conditions only on the observed data.
2.	Determine cluster(s) based on a threshold (critical t-value).
3.	Calculate the sum of the cluster(s).
4.	Take the cluster with the highest absolute value (Sum-score) if more 
than 1 clusters are present.
5.	Calculate the mean ERP activity, based on the highest cluster, in the 
individual trials.
6.	Use a non-parametric (Wilcoxon, Mann-Withney U, dependent of the 
type of experiment) test statistic to test whether there is a difference (two 
sided) between the two group means for this highest cluster.

Do you or anybody else think this is also a valid method for identifying and 
testing relevant ERP activity?

Best,
Emanuel

----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.