Rick Hass

One of the more concerning aspects of this is how long it will take for the new ballroom to be usable and what that says about how long Trump plans to be able to use it

Robustness is better assessed with a few thoughtful models than with billions of regressions | PNAS Robustness is better assessed with a few thoughtful models than with billions of regressions

Great piece on the absurdity of brute force multiverse analyses.

www.pnas.org/doi/10.1073/...

Chris Murphy: "There's a lot of history that has taken place in the East Wing & it was just destroyed without any conversation in the public, without any consent of Congress. It's absolutely illegal. That visual is powerful b/c you're watching destruction of the rule of law as those walls come down"

Oh wow

Frontiers | Validation of the Jefferson health-related social needs screener IntroductionScreening for health-related social needs (HRSN) is a growing national health priority. While multiple HRSN screening tools currently exist, none...

Just published in Frontiers in Health Services: we provide some validation evidence for a health related social needs screener in Jefferson Health. H/T to my colleague Sara Beachy for the clever use of latent class analysis #stats

www.frontiersin.org/journals/hea...

Most of the clinical prediction model literature seems to align with substantive expertise and correct model identification, and I believe there’s much more fruit to bear in this approach than in data mining AI pipelines

Tl:dr I’m still not convinced that engineering solutions to I’ll-defined prediction problems are the way we should be going in quantitative health research

So sure, you can get 1,000,000 “training examples” where say 10% are readmissions, but what are the key “features” for this problem? Should we simply pour it into an automated fitting procedure or do we get more out of simpler models with built with substantive insight on the causes of readmission?

Contrast that with trying to predict whether a patient will experience an unplanned readmission to the hospital in, say, 90 days from an index admission. If you ask physicians and nurses from 10 different specialties, you’ll likely get 10 different problem definitions.

I’ve been reading some old neural networks texts and one thing that strikes me is the vast difference between a problem like optical character recognition and that of clinical prediction models. The former problem being relatively well-defined with synthetic examples easily produced.

You’re not the only one, don’t worry!

Anyway Hakeem Jeffries and Chuck Schumer are both New Yorkers and the top Democrats in the House and Senate, respectively, and any point they could learn how to use a shutdown to make the point that we’re living in fascism.

They can keep holding press conferences in the Senate studio like every other press conference before it, talking to the same reporters, pretending like we live in a redeemable timeline where if they just keep on acting like it’s 2007, we might fall through a time portal that makes it so. But we live here in a country with a co-opted executive branch led by racist, xenophobic idiots who do not believe in law, empowered by a slavishly devoted Supreme Court that has decided shoveling unlimited power to said executive is more important than the law it disregards, and where the president thinks he can send national guardsmen from red states to blue states just because people of color live here and Fox News hosts hate that.

This part especially

I would think one fart = at least two seats 😂

I recently went for a Welcome Leap proposal on women’s health and couldn’t get over the use of the term “Thrust” instead of aim. It was especially off putting in that specific case 👀

Scene from Simpsons “The thing and l”: Hugo [evil Bart]: Look, I’ve been practicing. I made a pigeon-rat.

I might also recommend

“…there’s a little Uter in all of us.”

Remember folks. The original bluesky ethos by us bluesky elders (apparently I count?) is block and move on.

Don't quote dunk, don't pick fights.
Block, and move on.

This is what made this site unpalatable to the right early on, and we can continue to make it unpalatable to them.

Same thing is happening to Cross Validated, the StackExchange stats site, which I’m pretty sure has contributed to LLMs. Traffic and activity is way down, but often there are new questions like “I asked chatGPT … but I’m still not sure” ☹️

Hahaha! Yes!

Among the best 6 minutes in TV history

I know everything is awful but it’s that time of year again! 🎃 #Simpsons #Halloween

Yes, clearly you’re already a pro!

Pro-tip: well made New York bagels can be obtained in an approximate 60 mile radius around the city, just the same as NY 🍕

Yeah I was gonna chime in and say that squirrel and rat DNA have to be like 99.9% the same anyway

Trump found another $20 BILLION to give to Argentina, totaling their bailout to $40 BILLION.

Yet we can’t afford to prevent millions of Americans from losing their healthcare.

Shame.

Which is to say, all of us

At this point, between this garbage, the Argentina stuff and American Flag Swatstika guy (plus don’t forget ICE) I can say that I am worried that we’re only at the beginning of how awful this is going to get

Long story short, an actual Nazi doxxing ring. Enough people reported them that they went away for like 12 hours and then Elon personally brought them back. The site is dead. If you’re still on there, go lock your account right now

Important update: the Sprouts Market near me has cinnamon pecans in bulk that taste like Cinnamon Toast Crunch and they're gonna need to back a truck up to my house and dump about 2 tons of it or I'm gonna get real cranky