Another interest talk about doubly-robust confidence sequences: drive.google.com/file/d/1bAh0lX…

After a year of work, my review paper focused on predictive uncertainty methods for AI model interpretability is ACCEPTED! We hope to provide an accessible source for those unfamiliar with #XAI #Interpretability

Three presentations about #RiskLimitingAudits in Session 5 of the #EvoteID2021! The latest research on auditing complex #ElectionSystems was presented by @iwaudbysmith, Damjan Vukcevic & Michelle Blom, and @umbernhard. The session was chaired by Josh Benaloh from Microsoft Research.

This is such a huge honour! Also, thanks to my fantastic coauthors/mentors Philip Stark and Aaditya Ramdas (stat.cmu.edu/~aramdas/).

New review of incremental effects, with Matteo Bonvini, Alec McClean, & Zach Branson! arxiv.org/abs/2110.10532 Usual effects compare all vs none exposed - but this doesn't make sense if some are destined to be exposed. Can instead ask "what'd happen if trt were slightly more likely?"

I'm excited to share this paper I have been working on with my advisor Mikael Kuusela and collaborator Pratik Patil on UQ for wide-bin unfolding. Herein, we introduce a cool decision theoretic framework and use a prior without being Bayesian! arxiv.org/abs/2111.01091

Huge congrats to Amanda Coston on her Meta PhD research fellowship! research.facebook.com/blog/2022/2/an… Amanda's a joint student Machine Learning Dept. at Carnegie Mellon & Heinz College at CMU Carnegie Mellon University doing amazing work on fairness in ML, causal inference & more Check out her papers & more here: cs.cmu.edu/~acoston/

Is it possible to efficiently identify the optimal intervention while remaining agnostic to assumptions about the causal structure? In new work with Linbo Wang and Dan Roy, we study adapting to the presence of a d-separator using multi-armed bandits. arxiv.org/abs/2202.05100

The biggest downfall of conformal prediction has been its assumption of i.i.d. (exchangeable) data. Here, we develop a version of conformal that works with a form of distribution shift: feedback loops between the model and the covariates. Honored to work with seafann on this.

#ISSI This Thursday (8:30AM PT) we will have Neil Xu talking about “Post-selection inference for e-value based confidence intervals” (joint w/ Prof. Aaditya Ramdas & Prof. Ruodu Wang), followed by a discussion by Dr. Zhimei Ren Zhimei Ren selectiveinferenceseminar.com

Super excited to share new work "Comparing methods for statistical inference with model uncertainty" with Adrian Raftery that just got published in PNAS (PNASNews) ! #Statistics #DataScience #BMA (1/n) Paper: pnas.org/doi/10.1073/pn…

My advisor Edward Kennedy and I have posted an article on continuous trt effects estimation: arxiv.org/pdf/2207.11825… We study nonparam models where the dose-response curve has its own smoothness, which may differ from that of the outcome regression or cond dens of trt given X.

I am on the job market for faculty and research scientist positions. I develop statistical theory and algorithms to make data-based decisions that balance the need for robustness in high-stakes settings with strong performance in practice. Some highlights below: 1/n

Shantanu Gupta, Emre Yolcu and Minji Yoon of CMU School of Computer Science and Ian Waudby-Smith of Carnegie Mellon University Dietrich College have been named 2023 Amazon Graduate Research Fellows. cmu.is/fellows-23

Ramping up for this weekend’s #WiDSPittsburgh Carnegie Mellon University events! Come join us for #datascience talks and networking with Keynote by Sheryl Skolnick, Strategic Advisor UnitedHealth Group Optum! stat.cmu.edu/wids Registration is free! @WiDS_Worldwide Carnegie Mellon University Dietrich College

Join us in person or online for our Discussion Meeting next Tuesday, 23 May. Your comments are keenly invited on the paper ‘Estimating means of bounded random variables by betting'. Register for the meeting at rss.org.uk/training-event…

For the Stats crowd: unlike P-values, E-values (en.wikipedia.org/wiki/E-values) can be combined easily under arbitrary dependence. But can we say the same for their sequential analog, e-processes? Sadly the answer is *no*, motivating our recent work: arxiv.org/abs/2402.09698 (A thread)

Very excited about this new paper by Tiger Zeng (tigerzhzeng.com) We study causal inference w/ high-dimensional discrete confounders We give new bias/variance results & minimax lower bounds, which characterize fundamental limits of causal inference in high dimensions