Causal Inference and Graphical Models
Causal inference is a central pillar of many scientific queries. Statistics plays a critical role in data-driven causal inference. Jerzy Neyman, the founding father of our department, proposed the potential outcomes framework that has been proven to be powerful for statistical causal inference. Neyman’s framework has been influential in biomedical and social sciences. David A. Freedman used Neyman’s framework to critically examine many existing approaches for causal inference, and his work has enlightened several generations of statisticians.
The current statistics faculty work on causal inference problems motivated by a wide range of applications from neuroscience, genomics, epidemiology, clinical trials, political science, public policy, economics, education, law, etc. The faculty pioneer the principles, theories, and methods for causal inference building upon and extending the ideas from classical statistics (e.g., semiparametric theory, randomization inference, robust statistics), algorithms and principles from machine learning (e.g., random forest, stability principle), and optimization methods (e.g., evolutionary search and network optimization algorithms).
Beyond Neyman’s legacy of potential outcomes, the faculty also work on the theory of causal graphs that is relevant to practical causal inference. Nicholas Jewell co-authored a book titled Causal Inference in Statistics - A Primer, which is based on causal graphs.
causal inference in experiments and observational studies, with applications to biomedical and social sciences; contaminated data including missing data, measurement error, and selection bias
uncertainty quantification and inference, inverse problems, nonparametrics, risk assessment, earthquake prediction, election auditing, geomagnetism, cosmology, litigation, food/nutrition