Empirical Likelihood with Generative AI
Nonparametric Bayesian empirical likelihood methods for moment-restriction models, powered by AI-generated auxiliary data.
PhD candidate in Econometrics and Statistics
Jiguang Li
I am a PhD candidate in Econometrics and Statistics at the University of Chicago Booth School of Business, advised by Veronika Ročková.
I am on the 2026–2027 job market.
My research develops Bayesian methods for robust inference and adaptive decision-making, especially when relevant structure is latent, models are partially specified, or current decisions shape what data are observed next. Across these settings, I design scalable methods whose uncertainty quantification remains reliable when used to guide subsequent learning and decisions.
Previously, I was a full-time research professional at the Center for Applied Artificial Intelligence where I worked with Sendhil Mullainathan. Before Booth, I received an M.A. in Statistics from Yale University and a B.A. in Mathematics from Middlebury College.
Publications and Preprints
Dynamic Treatment on Networks
Offline reinforcement-learning methods for dynamic treatment allocation under network interference.
Robust Bayesian Optimization via Tempered Posteriors
Tempered posteriors for robust Bayesian optimization, with online calibration and regret guarantees.
Deep Computerized Adaptive Testing
Deep Q-learning for nonmyopic computerized adaptive testing with multidimensional latent traits.
Sparse Bayesian Multidimensional Item Response Theory
A scalable nonparametric Bayesian framework for learning sparse latent structure in multidimensional item response data.
Other Writing
Notes on Latent Variable Models
Notes based on David Blei's Build, Compute, Critique, Repeat: Data Analysis with Latent Variable Models.
Mixing Markov Chains
An expository note on fastest-mixing Markov chains on star graphs.
Undergraduate Thesis in Algebraic Combinatorics
The Chevalley-Warning Theorem: Its Proofs, Generalizations, and Applications.
On Exploration and Exploitation
RL case study.