Collaborative Research: Ranking of Entities

With limited resources available to tackle various challenging problems, policy makers and stakeholders often have to prioritize which problems to address in what locations or domains. In statistical terms, this involves the ranking of sub-populations and regions when limited (or no) directly observed data is available, and the available data can be on multiple aspects and of diverse types and modalities. The investigators will address the core theoretical, methodological and algorithmic challenges in such problems of ranking entities in multiple contexts of interest to the nation and to public life. The investigators will also develop techniques for measuring and quantifying the variability and uncertainty of such advanced, data-driven, principled ranking techniques to aid policymakers and stakeholders. For data to be analyzed using hierarchical models, subject to multiple sources of variability and dependencies, the investigators will develop reliable estimates of the ranks of entities with an appropriate quantification of associated uncertainty. The proposed methodologies will follow a Bayesian framework or a resampling-based frequentist one. While these techniques are primarily computation-driven, the investigators will address theoretical foundations of the proposed approaches both in the Bayesian and in the resampling-based frequentist paradigms. Using scalable computational techniques and leveraging geometric and topological properties of data, the investigators will also develop novel methods for ranking and identification of extremes when multivariate responses are of interest, and address benchmarking for compatibility over hierarchy of domains in a principled way. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.