Join us for a Quantitative Area Brownbag with Dr. Kate Hu (Ohio State University)!
Title: Using auxiliary information to adjust for confounding bias
Abstract: Observational studies often face confounding bias due to latent confounders being unadjusted. Modern open-data platforms offer unprecedented access to auxiliary data, yet standard methodological workflows underuse these resources. This talk presents a framework for leveraging negative control variables (NCVs), a type of auxiliary information, to adjust for unmeasured and mismeasured confounders, clarifying their underlying links to instrumental variables and proxy measurements. In the context of continuous exposures, I introduce a Bayesian nonparametric method to estimate a causal exposure-response function (CERF) while adjusting for unmeasured confounding using NCVs. By modeling the CERF as a mixture of linear regressions, the method achieves computational scalability without sacrificing flexibility for capturing nonlinear relations nonparametrically. Through extensive simulations, I show that the estimator successfully recovers underlying CERFs with various shapes despite unmeasured confounding. The session concludes with a practical demonstration in an environmental health study, showing how to select NCVs from auxiliary datasets to assess the causal impact of long-term exposure on cardiovascular hospitalizations in Medicare beneficiaries.
About Dr. Hu: Kate Hu is an Assistant Professor of Statistics at The Ohio State University whose research lies at the intersection of causal inference, study design, and semiparametric statistics. Her work develops innovative methods for drawing reliable conclusions from complex observational and environmental data, with applications spanning public health, climate science, and precision agriculture. Before returning to academia, she held leadership roles in data science at Aclima and Climate LLC, where she helped develop data-driven solutions to environmental and agricultural challenges. Dr. Hu received her Ph.D. in Biostatistics from the University of Washington and completed postdoctoral training at Harvard University.
