Stanford University · Department of Statistics
Assistant Professor of Statistics
David Huntington Dean's Faculty Scholar · Chamber Fellow
Statistical methods for distribution shift, causal inference, and replicability.
About
My research develops statistical methods for distribution shift, causal inference, and replicability; with the goal of making predictions trustworthy and scientific conclusions reproducible when the population that generated the training data differs from the one we ultimately care about.
I completed my Ph.D. in 2018 at ETH Zürich under Nicolai Meinshausen and Peter Bühlmann, followed by a postdoc at UC Berkeley with Bin Yu, Jasjeet Sekhon, and Peter Bickel. I joined the Stanford Department of Statistics in 2019. An applied strand of the work uses machine learning to help connect refugees with resettlement locations where they are likely to thrive.
News
Preprints & Publications
Augmented Inverse Hybrid Weighting: robust inference under deterministic and random distribution shifts
Which Covariates to Adjust for? Specification-robust Causal Inference in Observational Studies
Nonparametric Regression for Random Unbiased Perturbations
CTRL Your Shift: Clustered Transfer Residual Learning for Many Small Datasets
Optimal Empirical Risk Minimization under Temporal Distribution Shifts
Predicting data value before collection: A coefficient for prioritizing sources under random distribution shift
Out-of-distribution generalization under random, dense distributional shifts
Calibrated inference: statistical inference that accounts for both sampling uncertainty and distributional uncertainty
Beyond reweighting: On the predictive role of covariate shift in effect generalization
Learning under random distributional shifts
Tailored inference for finite populations: conditional validity and transfer across distributions
Model selection for estimation of causal parameters
Modular Regression: Improving Linear Models by Incorporating Auxiliary Data
The s-value: evaluating stability with respect to distributional shifts
Distributionally robust and generalizable inference
On the statistical role of inexact matching in observational studies
Causal aggregation: estimation and inference of causal effects by constraint-based data fusion
Anchor regression: heterogeneous data meets causality
Causal Dantzig: fast inference in linear structural equation models with hidden variables under additive interventions
Causal inference in partially linear structural equation models: identifiability and estimation
Guilt in voting and public good games
BackShift: learning causal cyclic graphs from unknown shift interventions
Confidence intervals for maximin effects in inhomogeneous large-scale data
Also: robust factor selection for biosimilar antibody production (Biotechnology Progress, 2017) and further work on economics & moral behavior. Full list on Google Scholar.
Mentoring
Teaching
Impact & Service
Online Causal Inference Seminar
I am a co-founder and co-organizer of the Online Causal Inference Seminar (OCIS), the leading online seminar in the field. Launched in 2020 during the pandemic, it draws roughly 50–70 attendees each week and has passed 350,000 views on YouTube — a platform for established researchers and early-career scientists alike.
Referee & editorial service
JRSS-B · Biometrika · JASA · Annals of Statistics · Operations Research · Biometrics · IEEE Transactions on Information Theory · NeurIPS (area chair, 2022) · UAI · Scandinavian Journal of Statistics.
Refugee resettlement
With the Immigration Policy Lab, Kirk Bansak (UC Berkeley), and Elisabeth Paulson (Harvard), I build data-driven approaches that connect refugees with resettlement locations where they are most likely to integrate successfully. The central statistical challenge is non-stationarity — both the economy and the characteristics of arriving refugees keep changing — so the algorithms are designed to stay robust to those shifts. IPL’s tools have been deployed in the United States, Switzerland, and the Netherlands, and the work was featured by Stanford’s Impact Labs.