[Eeglablist] ERSP baselining prior to condition comparison

[Downey, Ryan J]

The reviewer is mostly wrong but partially correct. EEG spectral power follows a 1/f profile roughly, with much more power at lower frequencies than higher frequencies. Thus, if you plotted the raw power as a function of time and frequency, you would see nothing particularly interesting (just some horizontal bars saying alpha has a lot of power for all time and gamma has relatively little for all time). Baseline correcting an ERSP adjusts for these naturally occurring differences in power across frequencies so everything can be visualized on a compact scale. So on the one hand you have to baseline correct or you won't be able to see the spectral fluctuations you are looking for; therefore the reviewer is wrong.

However, the reviewer is still partially correct in the sense that perhaps you are missing an important effect of the condition by removing the baseline power. It may be interesting for you to test for differences in the baseline power across conditions, in additional to comparing baseline corrected ERSPs across conditions.  Just note that there is a known scaling issue with newtimef such that the baseline corrected ERSPs are correct but the baseline power it returns is wrong. Someone made a modified version to report the correct numbers if you search for it on the eeglablist archive. Otherwise (what I'd recommend), you could just run a normal PSD calculation on the same [-500 -200] time window and that will be the baseline power that was removed. 

TLDR: The reviewer is accidentally partially correct. You should present both the baseline corrected ERSPs and the PSD (baseline power). Don't listen directly to the reviewer and simply get rid of your baseline corrected ERSPs altogether.


Ryan J. Downey
Postdoctoral Associate
Human Neuromechanics Laboratory
Biomedical Engineering Dept.
University of Florida



-----Original Message-----
From: eeglablist <eeglablist-bounces at sccn.ucsd.edu> On Behalf Of Eric Rawls
Sent: Tuesday, October 8, 2019 11:59 PM
To: eeglablist <eeglablist at sccn.ucsd.edu>
Subject: [Eeglablist] ERSP baselining prior to condition comparison


EXTERNAL EMAIL: Exercise caution with links and attachments.




Hi list,

I recently received reviews on a submitted manuscript that calculated the ERSP using the newtimef function in eeglab, with dB transformation using a baseline period of [-500 -200]. One of the reviewer comments is a little concerning to me, based on everything I've read. The reviewer writes (pasted from review):

"if I understand it correctly theta results are baseline corrected? It should not be when comparting conditions."

Based on everything I have read and several time-frequency workshop and courses I've taken, this seems incorrect. I think normally one wouldn't want to tell a reviewer they are incorrect in as many words, so I'm curious whether there's something I missed and how those of you experienced in use of time-frequency analysis might approach this.

Let me know what you think,
Eric Rawls, Ph.D.
NIMH T32 Postdoctoral Fellow
Neuroscience Department, Psychiatry Department University of Minnesota Medical School _______________________________________________
Eeglablist page: https://urldefense.proofpoint.com/v2/url?u=http-3A__sccn.ucsd.edu_eeglab_eeglabmail.html&d=DwICAg&c=sJ6xIWYx-zLMB3EPkvcnVg&r=Xqv3i1MFyNQV9dhuL1e9BtqA0SbF1UWmfPtBzxDT0xY&m=EhSPzCGtefvXjccZ-KwEMqm1bsAdsXpFgm_G5qpcvr0&s=ifhJruc5cAU35ZaWLW1DVFmTFx5j7XmTMRL46VaE-ik&e=
To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu