Dynamics at Scale: Statistical Frontiers in High-Dimensional and Streaming Time Series

1176 

Invited Paper Session 

Elynn Chen (1), Likai Chen (2), Elynn Chen (1), David Matteson (3), Yao Xie (4), Xiaofeng Shao (5), Likai Chen (2)

(1) New York University, N/A, (2) N/A, N/A, (3) Cornell University & National Institute of Statistical Sciences, N/A, (4) Georgia Institute of Technology, N/A, (5) Washington University in St Louis, Dept of Statistics and Data Science, N/A

Likai Chen  
N/A

Elynn Chen  
New York University

Elynn Chen  
New York University

David Matteson  
Cornell University & National Institute of Statistical Sciences

Yao Xie  
Georgia Institute of Technology

Xiaofeng Shao  
Washington University in St Louis, Dept of Statistics and Data Science

Likai Chen  
N/A

The explosion of high-dimensional, networked, and streaming data across finance, healthcare, economics, environmental monitoring, and the social sciences has created both extraordinary opportunities and profound methodological challenges for modern statistics. These complex data streams often involve dynamic dependence structures, structural breaks, and latent low-rank representations that defy traditional modeling paradigms. To address these challenges, statisticians have been developing innovative methods that integrate rigorous theory, scalable computation, and real-world applicability. This invited session brings together leading researchers at the forefront of high-dimensional and streaming time series analysis to showcase recent breakthroughs that advance both statistical frontiers and societal impact, in line with the JSM 2026 theme "Communities in Action: Advancing Society."

The session features five complementary talks:

Elynn Chen (NYU Stern): Tensor Recurrent Neural Networks with Statistical Low-Rank Structure Discovery – A framework for tensor-valued RNNs with provable low-rank structure discovery and uncertainty quantification, applied to finance, healthcare, and spatiotemporal forecasting.

David Matteson (Cornell): Network Modeling of Large-Scale Time Series with Cumulative Impulse Response Functions – A method for constructing sparse, directed networks from high-dimensional VARMA processes, enabling consistent impulse response estimation and insights into systemic risk.

Yao Xie (Georgia Tech): Online Kernel CUSUM for Change-Point Detection – An efficient kernel-based CUSUM procedure for rapid detection of structural changes in streaming data, with provable guarantees and real-world validation.

Xiaofeng Shao (WashU, St. Louis): Online Generalized Method of Moments for Time Series – An online extension of GMM offering efficient estimation and inference for dependent streaming data, with applications in stochastic volatility, IV regression, and anomaly detection.

Likai Chen (WashU, St. Louis): L2 Inference for Change Points in High-Dimensional Time Series via a Two-Way MOSUM – A multiple change-point detection framework using L2 aggregation and spatial-temporal scanning, supported by theory and applications to equity returns and COVID-19 cases.

Together, these talks provide a cohesive view of the emerging landscape of time series analysis in high dimensions and online settings, weaving together perspectives from tensor learning, networks, change-point detection, and online econometrics. The methods presented push beyond classical paradigms to address the realities of modern data: high volume, high velocity, and complex dependence. By unifying theory, methodology, and application, the session highlights how the statistical community is advancing the frontiers of dynamic data analysis while empowering societal decision-making in finance, healthcare, and beyond.

This session will appeal broadly to statisticians, data scientists, and applied researchers interested in methodological innovations for time series, scalable online inference, and the societal impact of modern analytics. The diversity of methodological approaches and application domains ensures a lively, cross-cutting discussion that will resonate across the JSM community.

Business and Economic Statistics Section ¹

Committee on Women in Statistics ²

International Chinese Statistical Association ³