🧵 redpill me on bayesian methods
Anonymous at Tue, 19 Mar 2024 22:49:41 UTC No. 16087196
I'm looking into a paper making use of some fancy bayesian inference model.
every time I wade through anything bayesian it will be:
* look at how beautiful my formula looks!
* oh, we can not actually estimate the prior / posterior, so let's use this approximation
* this is a trick to estimate another thing we can't know
* approximation of an approximation
* ...screeches in latex...
* ?
* claims profit without sensible baseline
also:
- every single bayesian statistics fan I met is a hopeless NPC
- just like haskell nerds, but without the programming skills
- bayesian memes are almost as bad as leftist memes
what do I need to know to defeat them?
Anonymous at Tue, 19 Mar 2024 23:57:51 UTC No. 16087298
Bayesians are at least honest when they're using approximations and prior knowledge. That said, I think frequentists might have an image problem because all the worst statistical work in science is done by people who just poorly apply the frequentist methods they learned in school/see other scientists use.
Anonymous at Wed, 20 Mar 2024 00:18:27 UTC No. 16087324