Concatenate in Coherence Analysis

Discussions
1

Hello,

I am calculating a coherence analysis with 30 subjects (2 groups, 6 conditions, resting-state-EEGdata). I chose Mscohere for my analysis. Every condition consists out of 30 artifact-free 4 sec epochs (= 120 sec continuous time interval) per subject. Even though I read the tutorial and various posts in the forum, it is still unclear to me, what exactly the process option "average cross spectra (one output file)" exactly does. Does it concatenate data first and then calculate coherence or does it calculate coherence per epoch and average the results. The number of windows ot the results indicate concatenating of the data. Since I have 4 sec epochs and I read in the forum, that coherence is best calculated with long continuous data, i was wondering if someone could give me a recommendation, wether 4 seconds can be considered as "long enough" or wether i should switch to a measure calculating coherence more accurate on short epochs (e. g. PLV) or concatenate data first.

Thanks a lot! :slight_smile:

Greetings
Christina

2

it is still unclear to me, what exactly the process option "average cross spectra (one output file)" exactly does. Does it concatenate data first and then calculate coherence or does it calculate coherence per epoch and average the results.

It computes the cross-spectra of for each input file separately, averages them, and then estimates the coherence from this average measure.
https://neuroimage.usc.edu/brainstorm/Tutorials/Connectivity#Coherence

This is the option you should use for the case you described here.