[Eeglablist] bootstrap-statistic ITC-values

Scott Makeig smakeig at ucsd.edu
Mon Sep 5 10:04:06 PDT 2005


Norman Habrich wrote:

>As a newbie to EEGLAB, i have a question concerning the ITC-values and their
>bootstrap-statistics. The ITC-values are covering a range from 0-1, where 1
>means absolutely inter-trial-phase-locked, o.k? During my calculations i set
>the p-values of the bootstrap-statistics to 0.05. The nice plots i get show
>significant time-frequency-ranges with ITC-values of 0.2. I would expect,
>that ITC-values below 0.5 could not be significant. How can i explain this
>significance?
>
>Thanks,
>Norman
>
>  
>
Norman -

The bootstrap statistical computation in timef() is straightforward - 
the assumption of time-locked phase consistency is broken by computing 
surrogate ITCs for collections of different latencies (see '>> help 
timef' and '>> timef details', as well as Delorme & Makeig, 2004, for 
details).

The ITC signif threshold is a function of alpha (e.g. p<0.05) and the 
number of trials. A signif. threshold of 0.2 could certainly obtain when 
the number of trials considered is relatively large.

Try using erpimage() with the 'phasesort' option to get some insight 
into the trial-to-trial variability that can produce an ITC ~ 0.2 ...

Scott Makeig



More information about the eeglablist mailing list