Laz

@Bayesprof Ah interesting. Generally, are there any researchers or even specific works you could point to that illustrate "building" in stats? I'm an incoming phd student and just trying to understand the landscape

@invictusqed no, not giving up on theory at all. There are various suitable analogues of "building" in theory as well, including playing around with algorithms numerically to design conjectures etc. But people are already doing some of this.

I don't know where PhD programs in statistics are going, but in the meantime, we can at least make the grad curriculum more useful: 0. Make gptpro/claude equivalent free for grad students; mandatory training and use of agentic workflow. Most universities still give only the 20$ version, not the most powerful one. 1. Revamp the grad programs and focus them toward "taste" and constructive criticism. I'd be ok with not having any written exams and instead having oral/take-home where students must use AI to replicate and then critique a published work to the committee's satisfaction. But here the standards have to be really higher than before for a student to "pass". This is the only thing we can still "teach". 2. Focus the grad program toward "building" instead of writing papers. A "thesis" can constitute constructing original data pipelines, assembling disparate data sources, open source of implementation of algorithms, etc. CS programs have been doing this for a while; stats has to catch up. 3. Formalize routine minimax arguments/convergence proofs via Lean, and focus on the key parts of the technical argument. In mathematics, up until recently, most graduate programs had a foreign language requirement. I wouldn't be surprised if they make Lean a language requirement in the near future.

@roydanroy How do you feel about the future of statistics research? Statistical theory is arguably mostly math, but it is quite a distinct flavor

The next era of mathematics will be owned by those who adapt. I'm not sure how long this era will last.

@visakanv @Thomasdelvasto_ Does that get gamed frequently?

@Thomasdelvasto_ in my country it’s a prerequisite for getting subsidised housing before you turn 35

I honestly don't understand why non-religious people even get married. Seems like nothing but a losing proposition to me

@Dorialexander What made you say this? Coming from an incoming phd in stats, thinking of going the bayesian route but doubting its current relevance😅

Forget about this, Bayesian is hot again.

@remilouf Great library! Do you keep up with the bayesian literature still, or strictly LLM space these days?

Context for those who missed my Bayesian inference years: github.com/blackjax-devs/…

Blackjax but for agent harnesses

Not sure if all this progress of AI for math is bullish or bearish for my phd.. guess we'll find out

@PradyuPrasad It's not even hot right now lol?

@__paleologo @mean_field_zane Besides Black Scholes which i think does fit this description, are there any other models in economics/finance literature that you think are good examples of modeling? With so many economists with a physics background, you'd expect the modeling style to seep through more often

*Models* trying to describe stylized should be simple to describe; they should be elegant and intuitive; they should have as few free parameters as possible; they should yield some good-to-very-good match with data. Mathematical rigor should be a secondary concern; in fact, the simpler the analysis, the better. Simulations should be welcome. This paper seems to be just a poor excuse to do some rigorous theorem proving. But it has no empirical value, it is not insightful, and it not very original or elegant mathematics. A lot of economic theory is like this. I am not sure what the value is.

The Obizhaeva-Wang model is a standard model of market impact (linear impact, exponential decay). For fun, I checked if I could solve the optimal MVO problem with AR(1) and alphas. I knew it is solvable but it’s gnarly to write the state-space eqs, the Riccati eq etc. A few prompts, a little thinking, some checking, and it was done in one evening, closed-form solution, LaTeX write-up and everything. ChatGPT Pro. It could be easily extended to AR(p) and multi-exponential decay. Human in the middle, but that’s some real time saving right there. So you can think about the actually original models. This is nothing special, but it would have changed the way I wrote my second book (📕) and its topics, had it been available in 2024.

@mean_field_zane @AgustinLebron3 @macrocephalopod @FangYi11101 Oof yeah that'll set you back a small grant :')

@invictusqed @AgustinLebron3 @macrocephalopod @FangYi11101 I need L3…

Can anyone inside the industry at a top place tell me how outdated this stuff is? Obviously academic empirical asset pricing folk are doing something different but I wonder how behind we are… academic.oup.com/rfs/article/33…

@mean_field_zane @AgustinLebron3 @macrocephalopod @FangYi11101 Daily data is not that expensive from something like databento is it? Hourly or higher freq does get quite spenny tho

@AgustinLebron3 @macrocephalopod @FangYi11101 CME sadly won’t share data with academics for free, just a 50% discount… has made a current project of mine stall until I can get serious funding lol.

@felpix_ @aadarwal @rieszspieces You need real analysis for political science phds these days?

@aadarwal @rieszspieces the political science ad coms will not be pleased with such a grade in analysis and a measly 168Q on the gre

analysis grade came back. never going to grad school

@avt_im @miniapeur Presumably this helps in ones career?

@miniapeur No, I mean why do you want 5+ papers?

Some people have really been triggered by this, especially in DMs. Guys, calm down, I just want to acquire the same superpowers, that is all.

@johngfriedman what about a stats phd?

If you are an international student with good tech options, recommend not doing econ phd. CS phd still good.

@macrocephalopod @mean_field_zane Are you saying academic finance does not have any value to you?

@mean_field_zane If academic finance is primarily concerned with variability in discount rates, and discount rates are equivalent to expected returns, and industry practitioners have (much) more accurate expected return estimates than academics, over every time horizon … what then?

I encourage all my QuantTwitter followers who shit on academic finance to read Cochrane’s AFA address. Academic finance concerns itself with very different topics than industry. I think most confusion arises bc industry types tend to read easy papers from crappy journals and not the more advanced stuff happening on the frontier, which is concerned with entirely different things than industry. static1.squarespace.com/static/5e6033a…

@shakoistsLog @mean_field_zane They're shitty fir prediction, but good for inference

@mean_field_zane i mean statistically beautiful. just never really work.

She’s so pure.

@mean_field_zane Will the econ department be seeing this money?

If you don’t understand how badly needed this is you’re NGMI.

@DimitrisPapail @ChenhaoTan How do you have it keep iterating then? Any task takes claude at most 30 minutes for me before it awaits a new prompt

@ChenhaoTan Pretty simple: Claude code (in vs code) + remote GPUs (ms sandbox or lambda)

Tenth night in a row that Claude code is running experiments for me overnight…

@alz_zyd_ What are good examples of this? I like pugh's analysis textbook which has these great picture proofs (and a guide on how to visualize 4D lol)

The best textbooks have rambly prefaces; weird long asides and personal anecdotes; random variation in content depth; furious handwaving and loose intuitions. Mathematicians try to purge their textbooks of these imperfections. But these imperfections are the soul of the work

Mathematicians like sterile textbooks. Textbooks distilled from their authors, the way Bourbaki is distilled from his various constituent humans. This philosophy is wrong. Every textbook is written by a human, and the best textbooks are unapologetically human

@JDHamkins @ben_golub @jayvanbavel Verification is probably easier than generation

@ben_golub @jayvanbavel Could you explain how this AI garbage is different from the AI analysis provided by Refine? I don't really see why I should think of them differently.

This is an existential terror manipulation for academics Watch AI create a complete, original job market paper in 6 minutes with almost no prompting.