[Eeglablist] Plotting CHannel Measures for STUDY

Brian Harvey brian.harvey at biogen.com
Fri Sep 6 09:00:31 PDT 2019

Hi team!

I am new to using this list and as I am using your tools more these days I am now thinking to reach out for some assistance, beginning with this...

I am the lead data scientist for an ERP (SSVEP) study at the VA here in Boston, MA and currently have recordings from 10 subjects across two repeat sessions. Each recording has 4 different visual stimuli randomly presented and stored in the eeglab structure as unique events (0Hz, 15Hz, 30Hz, and 40Hz)

When working with single session data, all ICA weights are applied within subject to all conditions and then I create seperate datasets from each event "type" afterwards. Clustering components works quite well to identify the steady-state evoked activity across subjects and comparisons across stimuli type work well (primarily against 0Hz, no flicker).

The issue comes when incorporating the second session datasets (essentially repeats of the experiment within subject on a different day). There is interest on the re-test reliability. Obviously ICA weights for these repeat recordings are different and thus ICA clustering fails for timef or itc measures. Arnaud explained that comparing across sessions using ICA clusters is not recommended.

Therefore I moved to channel measures for this... Precompute works great (although session 2 timef and erpimage files are written to the session 1 subject folders and I had to move them manually to the session2 subject folders)

Issue is when plotting these measures. When I choose a single channel for plooting ex. spectra looking at only session 1 vs session 2 (all events); as expected I get three subplots (Session 1/Session2 and t-test results). However if I chose multiple channels (all occipital leads) and choose to compare the mean of the channels for comparison it returns on one subplot.

This appears to be the case for plotting scalp topography of ITC for a specific frequency at a specific time...

Is this normal intended behavior?


More information about the eeglablist mailing list