Factor rotations & Varimax! Psychologists invented these ideas in the 1930s. Statisticians never liked it. But, if you ever needed it, there is now a statistical theory for factor rotations and varimax. published with discussion:

@mueblesfeos the people on X

@bennpeifert @SI27T Super interesting. The jumps smooth out in long run. Do you have extra terms to characterize jumps?

@bennpeifert @SI27T When does the fit break down / not work?

@bennpeifert @SI27T Roughly how many data points can you get for each fit?

@bennpeifert @SI27T statistician here... can you give us an idea (a thread?) on how you estimate these things, like the street smart version?

Great thread.

@kareem_carr Probabilities just vibes yo

RT @mattyglesias: The idea that identifying the people who run the government should be a firing offense does not make a lot of sense to me…

@TheStalwart Amara’s law!

RT @ben_golub: The US Congress is filled with cowards and failures, and I hope that people let them know it.

@ryxcommar @andimgladofit Maybe still enough to run R1 70B?

@ryxcommar @fedtanyl Banger

If you can’t tell the difference between two models, maybe you too have been ASI’ed

@analisereal @AdrianRaftery1

@ekernf01 yuup! my b! (someone just let me know on bsky too)

i'm sorry! this guy isn't "tank man"! this is a different guy that was at the protest

i got R1 running locally (70B). I asked: "can you tell me about the Tiananmen Square incident?" open source version doesn't have the censorship (which i'm sure others knew, but i didn't)

It will be like comparing manufacturing jobs in the 1950s USA to manufacturing jobs in the 1990s USA. It got displaced, some folks get rich. Folks in other sectors get cheaper stuff and no downside and wonder what the fuss is about.