Dwyer-Lindgren and UW Team Publish in Annals of Applied Statistics on Raking Mortality Rates Across Cause, Population Group, and Geography
Posted: 2026-10-30 12:40:00 (CSDE Research)
CSDE Affiliate Laura Dwyer-Lindgren (Health Metrics Sciences) and a team of UW co-authors from the Institute of Health Metrics and Evaluation (Ariane Ducellier, Parkes Kendrick, Bill Gustafson, Christopher Murray, Peng Zheng, and Aleksandr Aravkin) and the Department of Applied Mathematics (Alexander Hsu) recently published a research article in the Annals of Applied Statistics titled, “Raking Mortality Rates Across Cause, Population Group and Geography with Variance Estimation.” Raking adjusts observations in contingency tables to match known totals, reconciling estimates produced by models of different granularity. It is a key tool for the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD), whose outputs can be internally inconsistent, such as subregion totals that do not sum to regional totals. Variance estimation for raked values has typically required computationally intensive Monte Carlo methods. The authors propose an approach that achieves nearly the same variance estimates at the cost of a single solve. They also introduce a convex optimization framework that unifies several raking extensions, including differential weights, bounded estimates, feasibility checks, and handling of missing data. The authors illustrate the approach with a large-scale mortality estimation use case that reconciles estimates across race, county, and cause. The full article can be found here. Congratulations Dr. Dwyer-Lindgren and team!
Date: 10/30/2026