Research Overview

I develop quantitative methods for panel data, causal inference, and experimental design, motivated by decision-making problems in healthcare, digital platforms, and finance. My research has three connected themes.

Statistical inference for large-dimensional panel data

I develop latent factor methods that uncover patterns shared across units and time in high-dimensional panels, with an emphasis on inference, missing data, and interpretable structure. This work provides entry-wise inferential theory, with applications in causal inference and asset pricing. Topics include endogenously missing observations (JoE’23; Management Science, accepted), transfer learning across panels (JoE’24), state-varying dynamics (JBES’22), and sparse interpretable factors (JBES’22).

Representative Papers

• Large Dimensional Latent Factor Modeling with Missing Observations and Applications to Causal Inference [slides]
  Internet Appendix
  R. Xiong, M. Pelger
  Journal of Econometrics, 2023, 233(1), 271-301
  Honorable Mention, George Nicholson Student Paper Competition, 2019
  

• Factor Analysis for Large Non-Stationary Panels with Endogenous Missingness and Applications to Causal Inference [slides]
  J. Duan*, M. Pelger*, R. Xiong*
  Management Science, accepted
  

• Target PCA: Transfer Learning Large Dimensional Panel Data [slides]
  Internet Appendix
  J. Duan*, M. Pelger*, R. Xiong*
  Journal of Econometrics, 2024, 244(2), 105521
  

Causal machine learning

I develop causal machine learning methods for complex real-world settings. In healthcare, I study representation learning for multimodal clinical data with informative missingness across static modalities (EMNLP’25) and clinical time series (Findings of ACL’26). I also develop privacy-preserving federated causal inference methods for combining evidence across medical databases (Statistics in Medicine’23); methods for treatment effect estimation under network interference using semiparametric statistics (JBES, accepted) and causal message passing (NeurIPS’24; working paper); and stable prediction methods for unknown environments (KDD’18) and model misspecification (AAAI’20).

Representative Papers

• Causal Representation Learning from Multimodal Clinical Records under Non-Random Modality Missingness
  Z. Liang, Z. Pan, R. Xiong
  EMNLP 2025
  

• Federated Causal Inference in Heterogeneous Observational Data [slides] [code] [lay abstract]
  R. Xiong, A. Koenecke, M. Powell, Z. Shen, J. T. Vogelstein, S. Athey
  Statistics in Medicine, 2023, 1–22, doi: 10.1002/sim.9868
  

• Semiparametric Estimation of Treatment Effects in Observational Studies with Heterogeneous Partial Interference [slides] [talk] [code]
  Z. Qu*, R. Xiong*, J. Liu*, G. Imbens
  Journal of Business & Economic Statistics, accepted
  

Design and analysis of time-series experiments

I study how to design and analyze experiments that unfold over time, where treatment effects may persist and treatment timing is a central design choice. This work develops methods for choosing who should receive treatment and when, with the goal of reducing bias and improving statistical efficiency in applications such as digital platforms and public health. Topics include staggered rollout designs (Management Science’24; working paper), switchback designs (working paper), and automated design using historical-data simulations and gradient-free optimization (working paper).

Representative Papers

• Automated Experimental Design with Optimization from Historical Data Simulations
  R. Xiong

• Data-Driven Switchback Experiments: Theoretical Tradeoffs and Empirical Bayes Designs [slides]
  R. Xiong, A. Chin, S. J. Taylor
  

• Optimal Experimental Design for Staggered Rollouts [slides] [code]
  R. Xiong, S. Athey, M. Bayati, G. Imbens
  Management Science, 2024, 70(8), 5317-5336
  Finalist, MSOM Student Paper Competition, 2020