[Eeglablist] Advice for Abnormal Power Spectra?

[Kyle Swanson]

Greetings all!

Long time listener, first time caller.

For 4 data sets, I discovered when checking for 60Hz line noise that they
have abnormal power spectra -- They're almost like mountain ranges. At this
stage, I've imported channel locations, removed any channels whose
impedances wouldn't get low enough during the recording, filtered at .1Hz
and then at 100Hz, and then corrected for the delay introduced by our
amplifier. There's nothing odd in my notes from the recording session, and
I re-did these preliminary steps so I know that they've been done properly.
I also have many additional data sets that are perfectly fine.

I've consulted with my usual go-tos but I've had mixed feedback about what
to do. I've written up a little GoogleDoc
<https://urldefense.com/v3/__https://docs.google.com/document/d/1XpD64EL-YCQfuQo1scyyK306DQhHBITD8MBZWI0HV2s/edit?usp=sharing__;!!Mih3wA!UDouYXs1fr4eXzDq5M5P9UrHBzxOBIO1mrjiSz3FoYeI1Plogolf4ZbFFQZf3OtXcNw7JQ$ >
showing the issue and the results of some exploration that I've done.

Normally, after removing correcting any line noise using CleanLine, I'd
epoch the data and append all the trials into a single dataset, identify
and remove bad electrodes, identify and remove bad epochs, run ICA to
remove artifacts, re-reference, interpolate and removed channels, and
finally re-epoch so I have a file for each condition containing all trials
in that condition.

For this project, I'm looking at ERPs and ERSPs. So: Should I use these
data sets as they are? If not, any suggestions on whether I should try to
fix them and, if so, how -- especially in terms of integrating any remedial
steps into my pipeline as I've described it? Lastly, any ideas about what
the cause might be?

A big thanks in advance for any advice you might have!

Kyle Swanson

PhD Candidate
Department of Second Language Studies
Indiana University Bloomington
USA