From Data to Decisions Under Uncertainty: Imputation, Nowcasting, Integration, and Communication of Public Data

Government agencies increasingly rely on complex, large-scale, and often incomplete data to inform time-sensitive decisions. However, key challenges arise across the full data lifecycle: missing and inconsistent data require robust preprocessing and imputation, reporting delays obscure current conditions, emerging data sources demand scalable integration and modeling, and insights must ultimately be communicated effectively to diverse audiences. This invited session brings together four complementary perspectives on advancing data-driven decision-making under uncertainty. The first presentation focuses on multiple imputation for complex administrative data, highlighting system design considerations for handling heterogeneous, multi-source data. The second presentation addresses nowcasting and forecasting in the presence of reporting delays, comparing statistical and machine learning approaches for recovering recent trends from incomplete data. The third presentation examines scalable pipelines for transforming large-scale, high-frequency data into actionable insights, including integration and risk modeling. The fourth presentation explores how the public interprets data visualizations, using nationally representative evidence to inform effective statistical communication. Together, these talks span the full continuum from data cleaning and harmonization, to statistical modeling and prediction, to communication of results. The session emphasizes practical challenges, methodological innovation, and real-world applicability across domains such as public safety, health, and infrastructure. The session will include four invited presentations, followed by a discussant-led synthesis and audience discussion.