Rachel Leah Childers (@DonskerClass) header
Rachel Leah Childers (@DonskerClass) profile picture

Econometrics, Statistics, Computational Macroeconomics, Confusion. (Mostly the last.)

Zürich, CH
Joined June 2013
Photo by Rachel Leah Childers (@DonskerClass)Photo by Rachel Leah Childers (@DonskerClass)Photo by Rachel Leah Childers (@DonskerClass)Photo by Rachel Leah Childers (@DonskerClass)Photo by Rachel Leah Childers (@DonskerClass)Photo by Rachel Leah Childers (@DonskerClass)Photo by Rachel Leah Childers (@DonskerClass)Photo by Rachel Leah Childers (@DonskerClass)Photo by Rachel Leah Childers (@DonskerClass)Photo by Rachel Leah Childers (@DonskerClass)
For Spring semester, I'm bringing back free weekly open office hours for anyone in the world with Econometrics questions. Tuesdays 4-6 PM Central European Time (US EST 10AM-12PM) or by appointment; sign up and drop by!
2
2
20
1,797
With the rising tide of slop papers, I can no longer mainline arXiv, so I decided to fight slop with slop and vibecode a recommender system based on my library and own interests and papers to sort and select new ones that I would care about. Now live at: donskerclass.github.io/paper…
4
3
34
5,313
The format was inspired by @DrewDim's more ambitious personalized paper sorter: papers.ddimmery.com Sorting is mostly tf-idf with similarity comparisons to small embeddings. Code at github.com/donskerclass... though the hardest part was assembling my own library data.
1
6
470
Go see Emily to learn about principled ways to use synthetic data!
I’m at NeurIPS this week (12/2-12/8) to present our work on when/how synthetic data (e.g., LLM simulations) can help scientists make inferences with less real data, improving the efficiency of costly experiments. Come by Poster #904 on Thursday 4:30PM (Exhibit Hall C,D,E)!🙂
2
634
I will be at University of Tübingen on Tuesday Dec 2, and at NeuRIPS in San Diego Dec 3-8 spreading the good word of the Generalized Method of Moments in the age of AI. Econometrics lovers come say hi!
1
8
875
Updated Fall time for free weekly open office hours for anyone in the world with Econometrics questions. Wednesdays 3:30-5:30PM Central European Time (US EDT 10:30/EST 9:30AM) or by appointment; sign up and drop by!
1
3
4
973
Econometrics Free Open Office Hours details and signup at: donskerclass.github.io/Offic…
272
Valid inference w/ LLM-simulated data: 1. Take subsample of texts, extract vars 2. Construct moments identifying param on those vars 3. Ask LLM to simulate vars on sample & remaining texts 4. Use same moments w/ simulated vars 5. Combine moments, estimate jointly w/ 2-step GMM
💡Can we trust synthetic data for statistical inference? We show that synthetic data (e.g. LLM simulations) can significantly improve the performance of inference tasks. The key intuition lies in the interactions between the moments of synthetic data and those of real data
1
3
12
1,314
Working on this project to understand how LLM simulation data could contribute to drawing valid scientific conclusions from real world data, led by @yewonbyun_ with @shantanug7, was a great learning experience. Happy to see it forthcoming at #NeurIPS2025
1
325
For today's meeting, I read about Forward-Backward Stochastic Differential Equations. Among other things, these allow for a stochastic version of Pontryagin's maximum principle for optimal control along a path. Some notes I made explaining how they work: donskerclass.github.io/misc/…
Excited to kick off the Zurich 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠+𝐌𝐚𝐜𝐫𝐨 𝐅𝐢𝐧𝐚𝐧𝐜𝐞 𝐑𝐞𝐚𝐝𝐢𝐧𝐠 𝐆𝐫𝐨𝐮𝐩, organized with @DonskerClass & Felix Kubler! Faculty/students @UZH_en & @ETH_en will meet weekly to discuss both core CS papers and frontier economic applications.
4
15
3,146
For the Spring semester, I am restarting my free weekly open office-hours for anyone in the world with Econometrics questions. Wednesdays 10-12 AM Eastern or by appointment; sign up and drop by!
2
4
16
2,518
Link for Open Econometrics Office Hours details and sign-up: donskerclass.github.io/Offic…
3
490
New work by @shantanug7, @zacharylipton, and me on adaptively combining data sources for general semiparametric models. Think "design of experiments" but replace "experiments" with anything you can estimate by semiparametric GMM.
3
19
6,284
I'm restarting open Econometrics Office Hours this week, now from @BowdoinCollege, where I am visiting this year. Open to anyone anywhere needing econometrics help! Now Wednesdays 10-12 AM Eastern or by appointment.
1
38
1
205
36,570
"Fun" fact about threshold indicator rules of Sahm, Michez, etc type: they can incur unbounded regret against worst-case sequences, essentially since data can ping just above or just below an evolving threshold indefinitely. See S3.3 of cs.huji.ac.il/~shais/papers/… for the proof.
My latest column for the @FT analyses the various attempts to create "one simple rule" indicating that the US is in recession. It's... very difficult. 1st 300 clicks free to read... on.ft.com/4dM4ovQ 🧵🧵🧵
1
2
8
3,281
Against random sequences, such rules incur expected loss ~ O(d*Tmix/T) where d is number of predictors and Tmix/T is on the order of # of recessions (Yu 1994). This is "not enough data" if economy is β-mixing; in adversarial setting there's never enough. teapaw.org/SoumayaKeynes/status/1…
Replying to @SoumayaKeynes
My *bigger* concern is that we just don't have enough data points to do this exercise. h/t @Louis_Ashworth xkcd.com/1122/
1
2
950
I had a good time at @AcmSIGecom Economics and Computation #EC24 and apparently took ~25 pages of notes. Something to come back to after processing what I learned. donskerclass.github.io/misc/…
1
1
12
2,864
I'll be in NYC tomorrow, then New Haven for @AcmSIGecom #EC24 all next week, presenting at the Computational Methods for Economic Dynamics workshop. This is a new community for me, so I'm expecting to learn a lot. sites.google.com/g.harvard.e…
2
5
28
4,215
Office hours this week are moved to 9:00-11:00AM Eastern US due to scheduling. Potentially an opportunity for those in time zones where regular hours were inconvenient. Sign up: forms.gle/yhjPrQRNhpSAdhFx9
Trying something new: starting next Wednesday, I will be hosting weekly Econometrics office hours free and open to anyone in the world with Econometrics questions. Sign up at donskerclass.github.io/Offic…
3
1,321
The Double Pareto Log Normal distribution is commonly used for modeling economic inequality, but is hard to fit. I tried variational Parallel Tempering in Pigeons.jl and it seems to work well, so I'm sharing my tiny code exercises. github.com/donskerclass/Pige…
3
1,297
Computing pseudospectra today with pseudospectral methods. (These are near identically named concepts which are essentially unrelated.)
1
4
1,942
Happening again this afternoon, 4:00-6:00 Eastern! Sign up and link at donskerclass.github.io/Offic…
Trying something new: starting next Wednesday, I will be hosting weekly Econometrics office hours free and open to anyone in the world with Econometrics questions. Sign up at donskerclass.github.io/Offic…
1
3
9
4,290
For those with serious (statistical learning) theory questions, check out also this office hours initiative by a very impressive group. teapaw.org/let4all/status/1780399…
The Learning Theory Alliance is piloting a new office hours program! Drop in and chat about *anything* with experienced learning theorists, including @SurbhiGoel_ @thegautamkamath and Ellen Vitercik. Starts tomorrow at 9 AM with Gautam! More details in next tweet 👇
486
Thanks to participants in week 2! From next week I am opening up the video link. Please still send questions in advance via email or form if you want a considered response. donskerclass.github.io/Offic… teapaw.org/DonskerClass/status/17…
Trying something new: starting next Wednesday, I will be hosting weekly Econometrics office hours free and open to anyone in the world with Econometrics questions. Sign up at donskerclass.github.io/Offic…
2
5
2,510