We'd be grateful for any comments about points we overlooked, additional citations, as well as any corrections, clarifications, or suggestions for improvement! 🙏

We introduce metrics for quantifying the degree of alignment between the communication subspace and the dominant modes of input and output population activity. (e.g., Are the dominant modes of the input population the same ones driving communication?)

We also derive some useful (known) extensions, such as adding a ridge penalty ("ridge RRR") and non-spherical noise (accounting for correlated response noise), both of which preserve a closed-form solution.

Part of our motivation was our own difficulty understanding RRR and its mathematical origins (e.g., Why is this an eigenvector problem?). We thought others might benefit from a simple derivation and some figures and comparisons to build intuition.

Good find, Spencer! 🙌

This is excellent! 😂

In fact, Robbins 1956 ("An empirical Bayes approach to statistics") doesn't even consider Gaussian likelihoods. Only Poisson, geometric, binomial. So I'm puzzled about why this is the go-to citation. Are there multiple versions of this paper floating around??

Mind-boggling that HHMI would pull this right as other sources of funding for early stage investigators are drying up! 😢

Haha, ok thanks! 😂

Wow, great — thanks! I didn't know about CB education articles, but that could indeed be a good fit. It is indeed seeking to give technical — albeit (I hope) accessible — details of the derivations!

Ok, awesome — thanks for this suggestion! I didn't know about this journal before...

Cool — thanks for the suggestion! Although the web-page here "New Methods", which this paper is not. (Rather it's trying to give a clear, accessible description of an old method). Do you think that could fly?

Congrats, Takaki — this is super-cool!!

But there are no new results, per se. Any thoughts or suggestions for where to publish would be most appreciated!

The RRR estimator dates back to Izenman 1975, but we have found the original stats literature a bit hard to digest. So our paper paper aims to build intuition and give a simple derivation of RRR, along with several extensions (e.g., L2 regularization, non-isotropic noise).

By way of background: RRR is the method used for estimating a "communication subspace" between brain regions, introduced in Semedo et al 2019, and now growing in popularity for the analysis of multi-region datasets.

"Replacing animals with human-centered tools will provide better insight into human biology, speeding up the development of much-needed treatments for diseases like cancer and Alzheimer’s disease." 🤦‍♂️

The fact that the author refers to animal research as "animal testing" also reveals a lot.