Here's what I put together after several more days on the ground there:

margaretkilljoy.substack.com/p/our-neighb...

Here's a possibly relevant paper - journals.physiology.org/doi/full/10....

Nice! I would imagine that so long as there's limited driving noise, even waveform distortion (from filtering) won't affect an algorithm like this for burst detection - especially if rather than using just the median you ask for the IF dist to reflect the 1/f-normed PSD around the central freq.

Curious how this plays out for something that occurs as an AR(2) process as proposed for gamma… tends to have broader IF dist.

Also sleep EEG lit has ways to try and disambiguate bursts from non-bursts - esp. to find spindles which are right in that alpha range.
pubmed.ncbi.nlm.nih.gov/38650060/

Good point!

A naive question from someone looking more at brain oscillations - why not build avoid binning entirely and use generalized linear models where phase can be continuously modeled? If you have a multimodal relationship to the behavior (I.e. Hit rate) you use harmonies. elifesciences.org/articles/44287

For e.g. for patients with epilepsy, stopping seizures stops both the dynamics and clinical symptoms. That's the most drastic example I can think of. (3/3)

Think of the canonical example of gamma produced by PING.

So when seen in that light, to me anyway, if we can change oscillations, then we change the overall attractor state of the circuit. If we can do that, then it's inevitable we change function.

(2/3)