Patrick Perry

Patrick O. Perry is a statistician developing tools and methodology for nontraditional data, especially text and networks. He has worked on text summarization and scaling methods, dynamic network analysis, clustering methods for networks and other data, fitting methods for large-scale hierarchical models, and latent factor methods for high-dimensional data.

Perry's work has appeared in The Journal of the Royal Statistical Society, The Annals of Applied Statistics, and The Journal of Machine Learning Research, among other venues. He has developed and released open source implementations of his methods for the R software environment and has written a variety of other software packages for data analysis in the C and Haskell programming languages.

Currently, Perry is a Lead Data Scientist at Oscar Health. Prior to that he was an Assistant Professor in the Statistics group at NYU Stern teaching courses in introductory statistics, forecasting time series data, and statistics for social data. Perry received a BS in mathematics, an MS in electrical engineering and a PhD in statistics from Stanford University, and he completed a postdoctoral fellowship at Harvard University.

Research Interests

• Multivariate statistics
• Emerging data sources
• High-dimensional latent factor models
• Cross-validation for unsupervised learning
• Modeling and inference of network dynamics
• Social and economic networks

Academic Background

Ph.D., Statistics, 2009
Stanford University

M.S., Electrical Engineering, 2004
Stanford University

B.S., Mathematics, 2003
Stanford University

Companies

• Associate Director, Data Science Oscar Health (2018)
• Assistant Professor, Statistics NYU Stern School of Business (2011 — 2018)
• Postdoctoral Research Fellow Harvard University (2009 — 2011)