Pepe Montiel Olea

Econometrician at Cornell University. B.A. Economics/M.A. Economic Theory (ITAM, 2006); Ph.D. Economics (Harvard, 2013).

[Isaac Baley @isaacbaley]

Al final llegamos a Oviedo y la ceremonia fue un éxito! Gracias a todos :)

[Pepe Montiel Olea @PpM_O]

@isaacbaley ¡Muchísimas felicidades, mi querido (A)Isaac! ¡Qué grande! (Tenemos que rentar la Peter para celebrar como en la licenciatura)

[Pepe Montiel Olea @PpM_O]

A great example of the rapid advancement of science in our days. Courtesy of two of my favorite researchers: @lihua_lei_stat and @EdgarDobriban

[Lihua Lei @lihua_lei_stat]

New paper🚨After 10h w/ GPT-5.6 Sol, I made exciting progress on worst-case FDR of BH for correlated two-sided Gaussian tests

@edgardobriban showed FDR(BH_q)/q can exceed 1

Q (a folklore conjecture): Is the ratio universally bounded?
A: No👇

https://t.co/wILRtgZB7t

🧵(1/8)

[Pepe Montiel Olea @PpM_O]

@ElliotLip This is just a plain-vanilla multiplicative weights update routine, but it was kind of surprising to me this also works in the semi-infinite game you described. There is probably some better algo, because the approximately optimal strategy has 2M^2 ln(m)/epsilon^2 support points.

[Pepe Montiel Olea @PpM_O]

@ElliotLip The empirical distribution over P2's best responses is epsilon-optimal!

[Pepe Montiel Olea @PpM_O]

@ElliotLip (And if your intuition has some sort of computational interpretation on how to find the support points for the P2's strategy, you will be my idol)

[Pepe Montiel Olea @PpM_O]

@ElliotLip What a nice Sunday morning post! Would love to hear your intuition for this result. I was honestly very surprised about this when I first saw it. Theorem 2.4
2 in Blackwell and Girschik (54) has a version of this result for what they term "S games"

[Econometrica @ecmaEditors]

Local projections are surprisingly robust to misspecification. By contrast, VAR confidence intervals with short-to-moderate lag length can severely undercover for misspecification that is small, difficult to detect, and cannot be ruled out a priori.
https://t.co/rFObJ2dWfN

[Ashesh Rambachan @asheshrambachan]

Applications for the Machine Learning in Economics Summer Institute (MLESI) are open! Grad students, apply and come hang out in Chicago for a week to learn about ML and Econ.

[The Econometric Society @econometricsoc]

The Econometric Society: Announcing a new interdisciplinary conference on Economics and AI+ML https://t.co/k6pMfFn28s

[Pepe Montiel Olea @PpM_O]

https://t.co/joaCbedcxa

[Pepe Montiel Olea @PpM_O]

@CavaliereGiu @UCSD That is my favorite ECMA paper of all times! I wish I had met him! (And thanks for all your great posts!)

[Pepe Montiel Olea @PpM_O]

@TobiasSalz @SloanFoundation @ShengwuLi @instrumenthull @mdeshecon @guoxu_econ @alexoimas Great news, Tobi! Congratulations to you and the rest of this amazing group!

[Pepe Montiel Olea @PpM_O]

@RichRHC @dix_rafael ¡Muchas felicidades, mi Rich! ¡Te mando un gran abrazo!

[Pepe Montiel Olea @PpM_O]

@CavaliereGiu @JEconometrics Thanks for the tweet, Giuseppe! We had a lot of fun writing this! We started working on this project after reading this amazing paper from the late 80's: https://t.co/7tcMZTkatc

[Pepe Montiel Olea @PpM_O]

@fkocheng @LudvigsonSydney ¡Qué orgullo, Fede! ¡Felicidades! (Congrats to NYU as well for training such amazing scholars!)

[Pepe Montiel Olea @PpM_O]

@jmwooldridge @BruceEHansen [Extra] The posterior mean estimator corresponding to the Beta(\sqrt{n}/2, \sqrt{n}/2) can be shown to be minimax (under quadratic loss). From a decision-theoretic perspective, I find it hard to think about why to use any other estimator for this problem :)

[Pepe Montiel Olea @PpM_O]

@jmwooldridge @BruceEHansen 3) The posterior mean corresponding to any prior in the beta family can be written as a linear combination between the MLE and the prior mean. In my opinion, the prior that has a good finite-sample "frequentist" interpretation is the Beta(\sqrt{n}/2, \sqrt{n}/2).