Westat Experts Featured in JSM 2026 Proceedings

Westat researchers shared their expertise at the 2026 Joint Statistical Meetings (JSM) conference earlier this year, presenting on topics including emerging issues in survey methodology, artificial intelligence (AI), data quality and privacy, alternative data sources, health research, and science communication.

Proceedings papers from many of these presentations are now available, extending the conversations and research insights shared at the conference. The collection highlights innovative approaches to addressing complex methodological and analytical challenges, and demonstrates how Westat experts are helping shape the future of statistics, data science, and applied research.

Information on and links to individual papers follow. (Westat staff in bold).

Proceedings Papers

Addressing Reporting Challenges for NAEP Private Schools Through Multi-Year Data Integration

Tom Krenzke.

This paper explores a new approach to improving the reporting of private school results in the National Assessment of Educational Progress (NAEP) by incorporating matched schools from other assessment cycles. The findings suggest that this method can enhance sample representativeness and data quality, helping ensure more reliable insights into student achievement trends.

Expanding the Statistical Horizon: Integrating AI and Digital Tools for Cross-Disciplinary Innovation

Oluwagbenga David Agboola, Liz Mannshardt, Emily Kang, Matthew Lloyd Barker, Bochao Jia, and Zhang Haichao.

Drawing on the discussion from a JSM 2026 panel of experts from academia, government, and industry, this paper examines the integrated role of statisticians in the age of AI in framing problems, designing governance, ensuring data quality, and owning decisions and supporting them with evidence. The findings highlight how statistical expertise and rigor remain essential for ensuring responsible, trustworthy, and impactful use of emerging technologies.

Reaching Racial and Ethnic Minorities in an Address-Based Sample: How Much Efficiency Can We Gain?

Tien-Huan Lin, Daifeng Han, J. Michael Brick, and Xiaoshu Zhu.

Using data from two national surveys, this paper compares methods for oversampling racial and ethnic minority groups and assesses the impact on survey precision and efficiency. The results help researchers balance the need for reliable subgroup analyses with maintaining accurate overall population estimates.

Reducing Data Collection Costs for a Household Survey of Children: Leveraging Appended Variables on Address Sampling Frame

Daifeng Han and Sipeng Wang.

This paper explores methods that use auxiliary information for more efficiently identifying households with children in national surveys. By improving sampling approaches, researchers can reduce data collection costs and increase the effectiveness of studies focused on child populations.

Sampling and Weighting Strategies to Obtain Population-Calibrated Results From Nonprobability Survey and Multipathogen Respiratory Virus Serology Data Collected From Blood Donors

Laura Gamble, Elizabeth Eisenhauer, Andreea Erciulescu, Rebecca Fink, Sarah Smith-Jeffcoat, Jefferson Michael Jones, Bryan Spencer, Marion Lanteri, Eduard Grebe, Mars Stone, and Michael Busch.

This paper examines methods for building a nationally representative blood donor cohort to track respiratory virus infections in the United States. The resulting framework strengthens the use of survey and serologic data to support public health surveillance and preparedness for future outbreaks.

Sustainable AI: Deployment and Opportunities for Statistical Practice

Liz Mannshardt, Mahmut Kandemir, Cynthia Rudin, Steve Sain, Lyndsay Shand, and Elizabeth Eisenhauer.

This paper explores practical strategies for reducing the environmental and financial costs of AI systems while maintaining their effectiveness. It highlights the important role statisticians and data scientists can play in promoting responsible AI use through model selection, governance, interpretable machine learning, and efficient data practices.

Using Large Language Models for Missing Survey Data Imputation: A Proof-of-Concept

Hanyu Sun, Daifeng Han, and J. Michael Brick.

This paper evaluates the use of fine-tuned AI models to fill in missing survey data. The findings suggest that large language models (LLMs) can accurately reflect response patterns and survey logic, offering a promising new tool for survey research.