For a while now, @lauradee.bsky.social and I have been working on a guide to introduce Ecologists to causal inference tools for observational data that has problematic unobserved confounding variables lurking in the background. And now it's up as a preprint! www.biorxiv.org/content/10.1... - ThreadSky

jebyrnes.bsky.social • 361 days ago

For a while now, @lauradee.bsky.social and I have been working on a guide to introduce Ecologists to causal inference tools for observational data that has problematic unobserved confounding variables lurking in the background. And now it's up as a preprint!
https://www.biorxiv.org/content/10.1101/2024.02.26.582072v1

Comments

jebyrnes.bsky.social•361 days ago

In part, this comes from howmany of us have had "Correlation is not causation" drummed into our heads & abandoned or pushed aside projects w/ observational data. But other disciplines have greattechniques for causal inference w/ observational data!

So let's use 'em, Ecologists!

jebyrnes.bsky.social•361 days ago

The best part is that many of the techniques we show here are well within our grasp in terms of our training! We provide a code appendix (and all of our code is here: https://github.com/jebyrnes/ovb_yeah_you_know_me/ - see B for implementation)

jebyrnes.bsky.social•361 days ago

The worst part is that we show that the way we currently use mixed models with observational data for causal inference often produces a) the wrong answer and b) are underpowered. *sad trombone*

jebyrnes.bsky.social•361 days ago

We guide folk along the way with an example of looking at temperature impacts on intertidal snails and also provide some shiny apps to explore our simulations and results.

jebyrnes.bsky.social•361 days ago

But, the whole thing was an incredible journey, with @lauradee.bsky.social (SHE IS THE BEST) and I going up and down many rabbit holes, having a joyous collaboration, and laughing at how much it is clear we were both Witman Lab undergrads once upon a time.

jebyrnes.bsky.social•361 days ago

The paper is in review now (modulo some edits to word count), but we hope folk can pick this up now and make use of it ASAP. The methods are straightforward, more folk are using them, and have the power to really help us advance as a discipline.

https://www.biorxiv.org/content/10.1101/2024.02.26.582072v1

Comments

Posting Rules

Reply