[Eeglablist] Strange ERSP figures

Bingjie Cheng bingjie.cheng at uzh.ch
Fri Sep 3 06:26:14 PDT 2021

Dear EEGLABers,
I am working on EEG datasets of a navigation task in a virtual reality setting, and would like to look at spectral power changes when participants learned landmarks during the navigation. I used Bemobile preprocessing pipeline before entering the data sets in STUDY. 

I was checking the datasets by looking at the ERSP figures from STUDY. However, the ERSP figures at group level I got from the STUDY arevery strange. The figures have sharp state changes from positive values to negative values. One figure is attached here: ERSP at POz with 21 subjects. : https://urldefense.proofpoint.com/v2/url?u=https-3A__drive.switch.ch_index.php_s_1S2Pjlax3w3FSeJ&d=DwIFaQ&c=-35OiAkTchMrZOngvJPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=WlaYyqJo-qxCxQgQ8WE4Dcl4nNU2adLLOGeIrisngmo&s=6VDRlzOcN8Us9dJU4-85y5kNxP0FO55ReC80R5Ne5wo&e= 
 However, this problem doesn’t exist on the figure that computed at the single-subject level. The state changes are smooth on the  figures at the single-subject level. : Figure:  https://urldefense.proofpoint.com/v2/url?u=https-3A__drive.switch.ch_index.php_s_qbwfVpkJOPksaWF&d=DwIFaQ&c=-35OiAkTchMrZOngvJPOeA&r=kB5f6DjXkuOQpM1bq5OFA9kKiQyNm1p6x6e36h3EglE&m=WlaYyqJo-qxCxQgQ8WE4Dcl4nNU2adLLOGeIrisngmo&s=ggwyczCwQs47WVYGLNJSCQPCGDswuVh4KI9KWlrMCJQ&e= 

To summarize, as long as I group the subjects (no matter 10 subjects or 20 subjects), the boundary between positive and negative values become sharp, and this phenome does not exist at single-subject data.
Has anyone else encountered the same problem? Any idea why this happened?

More information about the software and pipeline:
- Matlab version: R2021a. EEGLAB version: v2021.0
- Analysis pipeline:
o Bemobile preprocessing pipeline: 
• Downsampled to 250 HZ
• Used ZapLine to remove line noise
• detected and interpolated  bad channels
• average reference
• ran AMICA to obtain ICs
o Removed eye componments
o Epoched data: There are in total 17 epochs (time window: [-1 5]) per condition per subject.
o Enter data in STUDY (21 subjects, 3 conditions). 
o Study design: categorical variable: condition – number of landmarks (3 vs. 5. vs. 7)
o Precompute channel measures, selecting ERPs, power spectrum, ERSP and ITC. I used the default parameters.
o Stats to plot the aforementioned ERSP figures:
• ‘Compute 1st independent variable statistics if any’ (-> in my case, 1st independent variable is the number of landmarks)
• ‘Use permutation statistics’, ‘Do not correct for multiple comparisons’, statistical threshold: ‘exact’, randomization: ‘auto’
Any help is appreciated! Thank you very much in advance!

Best regards,
Bingjie Cheng
Bingjie Cheng ¦PhD Student
University of Zurich, Switzerland

More information about the eeglablist mailing list