Rachel Leah Childers
LightNews LightNews
banner
donskerclass.bsky.social
Rachel Leah Childers
@donskerclass.bsky.social
2.4K followers 370 following 580 posts
Econometrics, Statistics, Computational Economics, etc
http://donskerclass.github.io
🇺🇲 in 🇨🇭. 🏳️‍⚧️
Posts Replies Media Videos
donskerclass.bsky.social
Never have I felt more like my job will soon by taken by AI. Statistical learning theory in Lean: concentration inequalities, Dudley's entropy integral, and local Gaussian complexity bounds.
30000 lines of code, over 1000 lemmas, formalizing Wainwright and Boucheron et al arxiv.org/abs/2602.02285
Fig 4 from Zhang, Lee, Liu "Statistical Learning Theory in Lean 4: Empirical Processes from Scratch" The dependency graph of the formalizations. Diagram shows proof of Dudley's entropy integral with preceding lemmas and Gaussian Lipschitz concentration likewise, feeding into a Gaussian Complexity inequality and an error bound for critical radius then used to prove sharp minimax error rates for linear regression.
February 3, 2026 at 7:20 PM
donskerclass.bsky.social
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!
Details and signup at: donskerclass.github.io/OfficeHours....
Free Weekly Econometrics Office Hours Email: rachelleahchilders@gmail.com or Sign up form: https://forms.gle/tSk48w1msZgrCUqX8 Time: Tuesdays 4:00-6:00PM Central European Time (GMT +1) (US Eastern Standard Time 10:00AM-12:00PM) (or by appointment) Location: Zoom Link https://uzh.zoom.us/j/62894466032?pwd=M28RFSrc44buAIJtPdvBEeYAIcClip.1 Who: Anyone. Grad students, researchers, government workers. Private sector is okay but in that case if your question requires work that exceeds the allotted time I may request to negotiate a consulting fee. What I can probably help with: Theory questions. Research design. Modeling. Particular expertise: Time series. Causal inference. Bayes. Structural approaches. Machine learning. Theory: Asymptotics. Statistical learning. Bayes/MCMC. Identification. Decision theory. Semiparametrics. Fields: I know most about macro (DSGE, heterogeneous agents, VARs, etc), can help with finance and applied micro (labor, development, health, etc), and can follow along with statistics problems in other areas. Code: I think in R, can write Julia, and can get by in Python. I am likely to suggest you build a model in Stan. I know Stata but if it’s relevant to your question I suspect you can get better help elsewhere.
February 2, 2026 at 3:54 PM
donskerclass.bsky.social
Peeking in on the old site, prominent macroecons are much more active there; they love multi-paragraph posts. Also AI-focused econs, which is less of a loss. Stats is still strong here.
But I was never really on Bsky for econ. What are you even doing on the mostly-Katie Tightpussy-by-volume site?
Profile header of Katie Tightpussy, Bluesky user @katie.bzky.team, highlighting a count of 236.6K posts.
January 29, 2026 at 9:45 AM
donskerclass.bsky.social
Huh; that delay differential equation has a closed form solution, and it's a perfect cosine curve under the right conditions. Not super surprising, but nice that it works out. doi.org/10.2307/1236...
From Larson (1964), "The Hog Cycle as Harmonic Motion", some key equations, derived from a continuous time hog cycle model with linear demand and in which quantities respond to prices linearly with a 12 month delay: (4) d/dt p_{t+12} = -bkp_t (5) d/dt q_{t+12} = -bkq_t If bk=1, this system yields a solution (7) p_t = cos(2πt/48 + e_p) (8) q_t = cos(2πt/48 + e_q) where e_p and e_q depend on initial conditions. If they differ by π radians, the solutions are consistent and the system is in resonance . Fig 1 from Larson (1964) "The Hog Cycle as Harmonic Motion" Relationships between the hog-corn price ratio, the numbers of sows farrowing, and pork production, as deviations from trend, 1947-1962. The chart shows real data for each series, overlaid with harmonic curves of differing amplitude and frequency. The visual fit is fairly good, up to high frequency noise.
November 18, 2025 at 3:05 PM
donskerclass.bsky.social
Ethel Cain puts on a world-class live show.
Photo of Ethel Cain and band on stage, dramatically backlit, surrounded by branches and hanging vines.
November 4, 2025 at 10:49 PM
donskerclass.bsky.social
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!
Details and signup at: donskerclass.github.io/OfficeHours....
Free Weekly Econometrics Office Hours Email: rachelleahchilders@gmail.com or Sign up form: hhttps://forms.gle/tSk48w1msZgrCUqX8 Time: Wednesdays 3:30-5:30PM Central European Time (GMT +1) (US EDT 10:30AM-12:30PM/9:30-11:30AM EST) (or by appointment) Location: Zoom Link https://uzh.zoom.us/j/64132987908?pwd=OzmlySLA3A3cjZz6RX7ibIuLoeJayU.1 Who: Anyone. Grad students, researchers, government workers. Private sector is okay but in that case if your question requires work that exceeds the allotted time I may request to negotiate a consulting fee. What I can probably help with: Theory questions. Research design. Modeling. Particular expertise: Time series. Causal inference. Bayes. Structural approaches. Machine learning. Theory: Asymptotics. Statistical learning. Bayes/MCMC. Identification. Decision theory. Semiparametrics. Fields: I know most about macro (DSGE, heterogeneous agents, VARs, etc), but can follow along in finance and applied micro (labor, development, health, etc). Code: I think in R, can write Julia, and can get by in Python. I am likely to suggest you build a model in Stan. I know Stata but if it’s relevant to your question I suspect you can get better help elsewhere.
October 28, 2025 at 5:40 PM
donskerclass.bsky.social
I've been at this a long time....
Tweet from Rachel Leah Childers, @donskerclass Favorite Banach spaces, ranked (partial order): W^k_2, Cameron-Martin space (any), L^2, L^∞, L^p (2<p<∞), W^k_p,..., L^1, L^p (1<p<2) 9:18 PM, August 27, 2015 Ranking does not necessarily reflect current views of the author. Apologies due to L^1 and their friends and loved ones....
October 4, 2025 at 9:23 AM
donskerclass.bsky.social
This post by Nan Jiang, nanjiang.cs.illinois.edu/2025/09/29/p... who knows more about RL theory than I ever will, on gaps in standard justifications for Policy Gradient, reassures me as I go through the PG lectures in interactive-learning-algos.github.io that its use is indeed a bit weird.
Office meme from Nan Jiang blog post Frame 1: Two cards reading "policy gradient with 100 tweaks" and "policy gradient with another different 100 tweaks", respectively. "Corporate needs you to find the differences between this picture and this picture." Frame 2: Pam Office: "They're the same picture"
October 1, 2025 at 10:17 AM
donskerclass.bsky.social
I was incredibly glad to catch #jasmine4t on tour last night! They put on an amazing live show, including new tracks from the extended album, out today!
Set list for jasmine.4.t's show in Zurich last night, given to me by the band afterwards Skin on Skin Tall Girl Breaking In Reverse Find Ur Pppl I Can't Believe I Don't Think Kitchen You Are The Morning Tesco Elephant Woman Did U No
September 12, 2025 at 11:01 AM
donskerclass.bsky.social
Meta-discourse point: under DeGroot learning, signals from multiple sources are reinforced, regardless of whether such information is downstream of a common source (see e.g. stanford.edu/~arungc/CLX....) "Tagging" signals appears to reduce this bias (Möbius et al 2015) but is rarely done.
Figure 7: Information aggregation in naïve model. From "Treasure Hunt: Social Learning in the Field" by Möbius, Phan, Szeidl https://www.markusmobius.org/sites/default/files/markusmobius/files/treasure_03_01_2015_0.pdf Lefthand figure: "No multiple paths" Diagram of directed graph with arrows C->B, B->A, D->A, with root nodes C, D highlighted. Arrows depict transmission of independent signals from B, C, and D to A, which in total are counted once each. Righthand figure: "Multiple paths" Diagram of directed graph with arrows C->B, B->A, D->A, and now also C->D with nodes C, D highlighted. Arrows depict transmission of independent signals from B, C, and D to A. Because both B and D are downstream of C and upstream of A, the signal from A contributes along both paths. Under naive (DeGroot) learning, signals from correlated sources are not discounted, so A "double counts" the signal from C. Identities of A, B, C, and D may be filled in by the reader.
September 5, 2025 at 3:59 PM
donskerclass.bsky.social
I’m pretty sure de basement is a primary input into the cruelty production process.
The Sisters of Dorley Welcome to Dorley Hall By Alyson Greaves Book cover. To spoil the joke, this is a book about a torture basement which I probably shouldn’t be mentioning to cis people.
August 15, 2025 at 3:58 AM
donskerclass.bsky.social
I haven't looked at the details, but this paper seems to have most of those? joelflynn.com/wp-content/u...
NBER WP 32602, the earlier version of the paper. Formerly titled "The Macroeconomics of Narratives", now "Contagious Business Cycles" Joel P. Flynn and Karthik Sastry Abstract: We study the macroeconomic implications of narratives, defined as beliefs about the economy that spread contagiously. In an otherwise standard business-cycle model, narratives generate persistent and belief-driven fluctuations. Sufficiently contagious narratives can "go viral," generating hysteresis in the model's unique equilibrium. Empirically, we use natural-language-processing methods to measure firms' narratives. Consistent with the theory, narratives spread contagiously and firms expand after adopting optimistic narratives, even though these narratives have no predictive power for future firm fundamentals. Quantitatively, narratives explain 32% and 18% of the output reductions over the early 2000s recession and Great Recession, respectively, and 19% of output variance.
July 31, 2025 at 9:46 PM
donskerclass.bsky.social
Gotta get on the next level (it's full of econometric theory)
Hand-size ringed ruled notebook with Rilakkuma (a Hello Kitt teddy bear) on it, with friends, dressed in a variety of nautical costumes including a seal, an octopus, a ray, and a shark. The contents of the notebook are primarily proofs related to the Generalized Method of Moments, and some derivations about Sobolev spaces.
July 30, 2025 at 4:37 PM
donskerclass.bsky.social
The title, "Debiased Semiparametric U-statistics" is meaningless to the average practitioner, but indicated it had exactly the theory I needed. It was hard to find because the title is now "Machine Learning Inference on Inequality of Opportunity" about 1 implementable estimator that uses the theory
The latest version of "Debiased Semiparametric U-statistics", with a new title: Machine Learning Inference on Inequality of Opportunity Juan Carlos Escanciano, Joël Robert Terschuur [Submitted on 10 Jun 2022 (v1), last revised 4 Oct 2023 (this version, v3)] https://arxiv.org/abs/2206.05235
July 19, 2025 at 1:39 AM
donskerclass.bsky.social
Papers are still coming out trying to fix it to this day. The client's setting was slightly nonstandard, so even those didn't work; doing it right would require combining multiple sources of theory, including results on U-statistics, another ugly area. Luckily, I found a paper w/ general results!
The paper I found on semiparametric U-statistics: Debiased Semiparametric U-Statistics: Machine Learning Inference on Inequality of Opportunity Juan Carlos Escanciano, Joël Robert Terschuur Submitted on 10 Jun 2022 (this version, v1), latest version 4 Oct 2023 (v3)] https://arxiv.org/abs/2206.05235v1
July 19, 2025 at 1:39 AM
donskerclass.bsky.social
Galaxy brain take: these are all splines. Polynomial is a spline with 0 knots. Categorical and binning are 0th order splines of different bandwidths. Covariate adjustment, CIs, and order selection can be done as standard in GAMs.
This is the philosophy behind binsreg: nppackages.github.io/binsreg/
Fig 4 from Cattaneo Crump Farrll Fung (2024) "On Binscatter" https://maxhfarrell.com/research/Cattaneo-Crump-Farrell-Feng2024_AER.pdf Panel A, B, C "Oversmoothed", "Optimal", and "Undersmoothed" show binned scatter plots with J=5, 11, and 50 categories, respectively, and accompanying best fit line.
July 16, 2025 at 1:37 PM
donskerclass.bsky.social
Rakesh Vohra is pretty good at titles
Bayesian Bullshit Sajan Srivastava Tymofiy Mylovanov Rakesh Vohra August 2024 Abstract A bullshitter neither knows nor cares about the truth, and therefore, it has been asserted, is more pernicious than a liar. We examine this assertion within the standard model of cheap talk communication where a bullshitter is modeled as an uninformed Sender. We show that in some circumstances, uncertainty about whether the Sender is informed or not can increase the welfare of the Receiver. Keywords: Bayesian Persuasion, Information Design, Cheap Talk JEL Classification: C6, D82, D83
June 28, 2025 at 4:29 PM
donskerclass.bsky.social
❤️
May 23, 2025 at 1:06 PM
donskerclass.bsky.social
Casey Corgi Childers
September 28, 2010-May 23, 2025
He was a very good boy.
Casey Corgi Childers, cuddling his pillow, earlier today.
May 23, 2025 at 12:45 PM
donskerclass.bsky.social
The new theoretical aspects of M-estimation w/ ∞-dim nuisances are mostly notation: same estimator & formulas, though with a generalized notion of derivative, under regularity. The actual challenge is performing the optimization. Also calculating derivatives, but bootstrap avoids that.
Screenshot of tweets: Quoting Tweet: I see that you have taken on a weighty vow. I respect it. tell the world. Quoted tweet (by me): PSA that due to long-ago choice of username, I am morally obligated to engage with any tweet regarding Donsker's theorem. Please use this power wisely.
May 12, 2025 at 9:19 PM
donskerclass.bsky.social
As the semester winds down I have a gap before student final papers come in in which to read general interest economics.

Anyway, Suresh Naidu, Winston Chen, & @treballen.bsky.social have fascinating work in progress on a quantitative GE spatial model of US slavery.
dl.dropbox.com/s/uvhbdrqr88...
Estimated productivities, amenities, and coercion Maps of the US South c1860 with county-level values from an estimated structural model based on population and production data. Of the 3, the most visually notable is the estimated coercion, a measure of the gap between enslaved and free worker populations, which concentrates in the cotton belt and around the Mississippi. Results in the presentation show amenities correlate negatively with malaria, while productivity shows weak correlation with cotton production, which should explain the lack of geographic concentration, How does emancipation affect the allocation of labor? Maps showing the equilibrium spatial allocation of formerly enslaved and free population in a counterfactual exercise in which slavery and coercion are completely removed in the US South in 1860 (probably not a fair comparison to actual emancipation due to continued coercion and restrictions on mobility post-bellum). According to this model formerly enslaved people would relocate from the cotton belt especially to Missouri, Appalachia, and Florida and formerly free people would make the reverse change. I imagine historians with regional knowledge would have plenty to say about the specifics.
May 3, 2025 at 8:20 PM
donskerclass.bsky.social
I don't know; I'm pretty happy to let non-specialists come to me and learn to understand what I know starting from first principles, with time and effort.
Whether this counts as "accessible" might involve a discussion with our Office of Student Aid, though.
Table of Bowdoin 2024-2025 Cost of Attendance Direct Costs Tuition $67,198 Fees $634 Housing $9,148 Food $9,340 Direct Costs Total $86,320 Indirect Costs Books, Course Materials, Supplies and Equipment $840 Miscellaneous Personal Expenses $1,660 Transportation varies Indirect Costs Total $2,500+ Total Cost of Attendance $88,820+ This page is meant to be an overview and an introduction to costs and aid at Bowdoin. When you're ready to talk about detailed numbers and processes, the Office of Student Aid employs five of the best listeners we've ever met—they work with each applicant's family, one-on-one.
May 1, 2025 at 7:35 PM
donskerclass.bsky.social
Oh, thanks, I completely missed that! Expect more vibrant threads on Bayesian updating from here on out.
April 13, 2025 at 3:21 PM
donskerclass.bsky.social
Not saying, but...
Screenshot of tweet by user kassanove @deuxnan *** Call Me By Your Name was the first socially acceptable movie for fujoshi's to hit the mainstream. i think Sufjan Stevens bridged the divided between girls on tumblr who stink bad and the judges at Cannes. Timothee Chalamet has live action anime adaptation yaoi boy physiognomy
April 5, 2025 at 6:28 AM
donskerclass.bsky.social
There are some very unique suggestions about unconventional sources for a medical literature review in this episode.
Header for Archive of Our Own followed by search results: "You searched for: back pain sort by: best match descending 22,199 Found"
March 31, 2025 at 12:48 PM