Federal Grant Award Summary

The National Science Foundation's Division of Mathematical Sciences awarded $146,715 to the University of Michigan under the Mathematical and Physical Sciences program (CFDA 47.049) on June 1, 2026, for a collaborative research project on calibrated hypothesis testing. The project, scheduled for completion by May 31, 2029, will develop statistical theory and methodology to ensure that error rates reported in scientific findings are accurate and interpretable. The research addresses a critical gap in current multiple testing procedures by extending traditional false discovery rate methods to provide individual findings with calibrated error probabilities rather than only controlling average error rates across entire discovery lists.

The deliverables from this research initiative include the development of novel statistical frameworks combining empirical Bayes estimation with frequentist finite-sample guarantees, application of these methods to variable selection problems using knockoff and sign-symmetric statistics, and extension of local and boundary false discovery rates to settings with dependent p-values. Additionally, the project will produce open-source software tools, graduate student training in advanced statistical methodology, and new instructional materials to support broader adoption of calibrated hypothesis testing in scientific research and data-driven systems. These outputs are intended to enhance reproducibility in large-scale scientific studies and improve safety in artificial intelligence applications where suspicious inputs must be systematically triaged and evaluated.