Jason

computational math / stats (bayes). @getrecast.

[Chelsea Parlett @ChelseaParlett]

"you need to correct for family wise error rate" 🐹🗣

#pikasShoutStats

[typedfemale @typedfemale]

i definitely understand why austen allred wouldn't like a tool that reports you to regulators if you commit fraud

[Vincent Woo @fulligin]

I'm launching https://t.co/HWDYjbBeY4 to celebrate the actual humans who have served as the coal that we've shoveled into the AI GAN engine. These 70,000 people exist. You exist. I exist.

[Jason @JasonPekos]

@bschrdr07 @Duderichy Even if I accept this framing, it sounds like TD bank specifically was helping to facilitate the drug trade, and they were appropriately fined?

Which seems like a much more appropriate mechanism than diffuse tariffs.

[Jason @JasonPekos]

@just_cameron saddest moment of my life: finally moving to an american tech city (seattle), only to realize that we still dont have waymo

[Jason @JasonPekos]

@nominalthoughts I think the worst part was that it required enough focus that I couldn’t zone out and listen to a podcast or something. Just endless hours of focusing on poorly written solutions to the same three calculus problems – terrible!

[Jason @JasonPekos]

@nominalthoughts I really really really hated grading for first year life science calculus (as a TA) — hundreds of tests, tight grading deadline, and so much random BS for students. Grading for smaller courses wasn’t that bad.

[Jason @JasonPekos]

@pli_cachete Yup, I did the same thing last year and came to the same conclusion. I think this was also clear from the (lack of) response to @natalia__coelho …

[Jason @JasonPekos]

@ed_kroc reminds me of

[Jason @JasonPekos]

@nominalthoughts honestly hes the lowest variance politician ive ever seen. even here, you listen to what hes saying, and really he wants longer approval timelines, more stringent regulation on cereal or whatever.

[florence ⊮ @morallawwithin]

Nate Silver is gonna be so embarrassed if it turns out that one of the candidates had a 90% chance of winning

[Jason @JasonPekos]

also, like, I took measure theory too. I dont think it provides any insight into the statistics wars. Im fine being a pragmatist. Its fine, the models works. It’s fine. but the idea that “knowing math” alone has relevance to this question seems crazy to me.

[Jason @JasonPekos]

the roomba doesn’t care that the interpretation of probability it uses to do location detection is “generally meaningless” – it just works.

[Sridhar Ramesh @RadishHarmers]

I understand math and I basically agree with this. Bayesianism is misguided and the application of the mathematics of proportion (aka, "probability") to modeling the uncertainty around a sui generis event like the election is generally meaningless.

[Jason @JasonPekos]

the funny thing is that nate silver is also not really much of a bayesian, as wasserman points out

[Jason @JasonPekos]

also goes for SMC

[Jason @JasonPekos]

I feel terrible running my, like, eight serial mcmc chains while the neural net gang gets to massively parallelize everything on their fancy gpu clusters. there’s got to be a better way, right?

[Jason @JasonPekos]

I’ll believe in VI when I see it

[Jason @JasonPekos]

@ChelseaParlett @just_cameron Was just looking at a place a few days ago that had two rooms like that + a bedroom. listed as 3br!

It’s not a bedroom if you can’t fit a bed in it!

[Jason @JasonPekos]

@just_cameron But how consistent is it?! 4o seemed to be able to help me with tricky measure theory stuff, maybe, like 10% of the time.

The rest of the time, it made totally embarrassing mistakes.

There’s a lot of hype — I’m not convinced this isn’t just people screenshotting the successes!

[un-monitored, @randy.pub in the good place @randyboyes]

New TidierPlots theme made for the website, along with an explainer if you want to make your own Makie/TidierPlots themes: https://t.co/Qr8AKLqhKQ