Professor Kuffner's research spans statistical theory, foundations, and methodology.
Using techniques from asymptotic theory, Kuffner studies the questions of validity, accuracy and power of statistical inference procedures, and also develop methods which can help achieve these goals. He is interested in the relationships between different paradigms for statistical inference, such as neo-Fisherian, Bayesian, and frequentist approaches, and also new methods for data-driven science from machine learning. In addition to ongoing work on higher-order asymptotics for likelihood-based and Bayesian inference, he is currently working on post-selection inference, the bootstrap, and prediction after model selection.