Mauricio Caceres Bravo

My intuition is that if the true rate is 98% then the upper bound for discrimination is 4 percentage points, right? So grand jury discrimination shouldn't be able to explain more than 4pp of the felony conviction gap downstream, or am I missing something?

Can you look at media coverage of inflation? Here it would include tariffs.

Presumably they misunderstood the middle half to be the "median scores" (the "median half"). My impression is that 25% to 75% is an oft reported range in this context.

The paper estimates 0.1: Small overall but in line w/elasticity of locals (0.03 to 0.15 from a cursory glance), which is the entire theory they try to test.

Not familiar w/lit but also notable all 4 papers study broader income distributions (of narrower populations) than all US millionaires.

I would earnestly be quite interested in the evidence you have that's convinced you "unpopular cultural issues" cost Ds the 2024 election.

I've my own thoughts on this but I've not done anything like a deep dive.

Not if you're discrete about it.

Couldn't we also eliminate the nickel, then?

I think even now if you look through headlines for many publications you wouldn't necessarily get the impression that what happened and continues to happen is criminal.

Court rulings are also often described as if enforcement was automatic, further obscuring what's actually happening on the ground.

The idea is laughable on its face, but even if we grant them their premise, they completely ignore the irreversible suffering they can cause while treating our country like a college project.

I had an experience like this in the US for my first "high-skill" job. During the interview it came out that I'd worked a "low-skill" job for two years (I hadn't put it on my resume). My future boss became way more interested, and explicitly said that was key to seal the deal.

Musk's Junta Establishes Him as Head of Government Imagining how we'd cover overseas what's happening to the U.S. right now

www.doomsdayscenario.co/p/musk-s-jun...

Yes, the dynamic started with everyone focusing on tariffs while Musk secured control of the Treasury and killed USAID, and has been ramping up since.

Well, you need to assume that whatever caused the accident isn't a feature of the place that can make accidents more likely, which needn't be true.

I got the font I already use in my shell/text editor (Source Code Pro). Were you already using Fira Mono?

That's interesting. I learned early on that // and /**/ don't work interactively and moved on, but you're right that it could catch someone out.

I guess * being a comment in Stata but dereferencing a pointer in mata would be another one?

I see. I was thinking about it more, and I've also been caught out by this on occasion, actually, and I've made assumptions based on past experiences with other languages.

Ultimately it just comes down to different ways different programs work.

Actually, here's one I hadn't thought about, and maybe was a "gotcha." When u hit "run" in Stata's editor it makes a temp file that runs via "do," but when I first started I expected it to run via "include" (bc that's essentially what happens in SAS when u hit run). Had a few mishaps with locals!

Oh, you meant "instead of"? I guess I don't usually think of Excel as competing with R/Stata/SAS but as a different program with specific use cases.

Tho if I think about it, I did learn regressions in Excel in college.

I was just curious what people put forth as examples of "gotchas" in R or Stata. I don't know that I perceive what people may call unintuitive or surprising between them so easily because of the timing of when I learned each program.

I thought everyone could benefit from learning spreadsheets. Why is it bad?

When I learned Stata I already knew a few other languages so I don't think I appreciated what was supposed to be easier or harder about it at the start.

I'm genuinely curious: What are these gotchas that people talk about with R and Stata?

Great thread!

Another tip: gcollapse has a -merge- option that basically does multiple gegen calls at once (the original idea was to mimic collapsing multiple variables and merging back the collapsed data, hence the name).