YSPH Biostatistics Seminar: “Cross-study blending for generalizable prediction of UFP concentrations"

NOTE: BIS 525 students are required to attend in person. Others are invited to attend in person, but may also attend via Zoom.

SPEAKER: Prasad Patil, PhD, Assistant Professor, Department of Biostatistics, Boston University

TITLE: “Cross-study blending for generalizable prediction of UFP concentrations"

ABSTRACT: Atmospheric concentration of ultrafine particles (UFP; particles <100nm) has the potential to affect human health in a distinct manner from larger particles such as PM2.5. In a study of aviation-attributable UFP from flight activity at Boston Logan Airport, our team implements stationary monitoring at varying distances along flightpaths from the airport. One goal is to develop a predictive model for UFP using meteorological and air traffic measurements that is generalizable to regions where direct monitoring cannot be conducted. Due to variations in orientation to the airport and local environmental factors, UFP measurement distributions from our monitoring sites are heterogeneous. There are advantages to training prediction models on the combination of all site data or by weighted combinations of predictions from single-site models. We show that for unbiased models, an optimal blending of predictions from both approaches performs at least as well as each scheme adaptively at differing levels of heterogeneity. We apply insights from these findings to propose an empirical weighting strategy based on cross-site model performance and agreement.

YSPH values inclusion and access for all participants. If you have questions about accessibility or would like to request an accommodation, please contact Charmila Fernandes at Charmila.fernandes@yale.edu. We will try to provide accommodations requested by October 2, 2025.