[Eeglablist] toolboxes for multiple testing of time-frequency data

cyril pernet cyril.pernet at ed.ac.uk
Thu Jan 23 01:42:55 PST 2014

Hi Arnaud,

> According to Guillaume Rousselet and Cyril Pernet, this is the only 
> method that adequately control for type I error rate when correcting 
> for multiples comparisons (so if you have an uncorrected threshold 
> p-value at 0.05, you corrected p-value threshold will be exactly 0.05 
> corrected for multiple comparisons - which might not be the case when 
> using other methods). I tend to agree with them.

In fact, in our LIMO EEG paper we show that F max also correct well, and 
in fact so does TFCE (in prep - will also be presented at OHBM this year).
So max statistics, cluster-mas, and TFCE are three methods which 
provides a good control over the family-wise type 1 error rate, that is 
at e.g. p=0.05 you will have about 5% false positives under the null 
hypothesis (H0). All three techniques however do differ in terms of 
power (i.e. under H1) with cluster-mass better than stat max, but at the 
cost of losing inferential power as cluster-mass only allows drawing 
conclusions about clusters. For TFCE it is not clear yet, but some 
results show it performs just as well as cluster-mass, and provides 
inference at the cell (time frame - electrode) level.

All 3 methods are available in the LIMO EEG toolbox

Dr Cyril Pernet,

Academic Fellow
Brain Research Imaging Center
Division of Clinical Neurosciences
University of Edinburgh
Western General Hospital
Crewe Road
Scotland, UK

cyril.pernet at ed.ac.uk
tel: +44(0)1315373661

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

More information about the eeglablist mailing list