Theory related to prediction modeling
Our theoretical work focuses on the fundamental ingredients of prediction—especially how uncertainty shapes what we can (and can’t) reliably predict. Prediction modeling isn’t about explaining the past; it’s about making solid forecasts for the future, and that requires a different way of thinking about statistical models. We work on theory that clarifies how predictive performance behaves under realistic conditions, including messy data, shifting populations, and imperfect models. A big theme is understanding uncertainty in all its forms—because every prediction comes with some level of “unknown broth.” By developing theory grounded in real-world challenges, we aim to better define the limits of prediction and provide a stronger foundation for everything else we do in the lab.