Hello, I’m analyzing EEG data and want to compute connectivity (NxN correlation matrices) between current amplitudes of regions of interest (ROIs) in the source space during a task period (~20 s). I plan to compare the connectivity between an experimental task and a control task.
I can compute the connectivity with two ways and would appreciate advice on which is more appropriate:
- On the node of the already averaged sources across all trials, then compute the NxN correlation matrix from these averaged time series.
- Or take all the trials in the Process space and perform connectivity there by selecting average all.
The two approaches give me different results. Which one is the most appropriate?
Thank you in advance!
Which is more appropriate depends on your research question, because the two approaches measure slightly different concepts.
-
Connectivity of trial average: it measures the connectivity only for the phase-locked (evoked) activity, as the non-phase-locked activity (even if it is time-locked) it cancelled out in the trial averaging.
-
Average of trial connectivity: it measures the connectivity for the phase-locked and non-phase-locked (evoked and induced) activity. This happens as for each trial, its connectivity metric describes both evoked and induced response, so the average across trials reflects both responses.
In terms of methodology, it is very similar to the computation of PSD, it can be done as:
- Power spectrum of average (describes evoked power), or
- Average of power spectra (describes evoked+induced power)
2 Likes