Mitigating spatial confounding in observational studies
Exploring methods for addressing biases from unmeasured spatial confounders, helping to improve the accuracy of causal estimates in environmental health research.
![Logo for the Centre for Data and Statistical Science for Health on a blue background](/sites/default/files/styles/event_main/public/dash-centre-event-card_0.jpg?itok=td3c0tS1)
As part of a seminar series in environmental epidemiology, Professor Brian Reich from the Department of Statistics at North Carolina State University will discuss approaches for mitigating spatial confounding in observational studies.
His talk will explore methods to address biases introduced by unmeasured spatial confounders, such as leveraging spatial smoothing and propensity score models. Additionally, he will highlight case studies demonstrating how these techniques improve the accuracy of causal estimates in environmental health research.
This seminar series, part of DASH’s Improving Environmental and Planetary Health challenge, will use real-world examples and critical discussions to explore challenges in assessing causality in environmental epidemiology, and highlight innovative approaches to improve public health insights.
It will consist of three seminars given by speakers from Harvard Medical School, Harvard T.H Chan School of Public Health and North Carolina State University, followed by a fourth internal LSHTM panel discussion event.
Talk abstract
Environmental studies are often observational and processes of interest exhibit spatial dependence, which presents challenges in estimating the causal effects. Spatial data can also be subject to preferential sampling, where the sampling locations are related to the response variable, further complicating inference and prediction.
To address these challenges, we propose a spatial causal inference method that simultaneously accounts for unmeasured spatial confounders in both the sampling process and the treatment allocation. We show that the causal effect of interest can be reliably estimated under the proposed model, and apply the method to estimate the policy impact of marine protection areas.
Speaker
Brian is the Gertrude M Cox Distinguished Professor of Statistics at North Carolina State University.
His research interests include Bayesian methods, spatial statistics, extreme value analysis, variable selection and dimension reduction and machine learning. In addition to these methodological interests, Brian applies these methods to environmental areas such as ecology, epidemiology, meteorology and climate.
Event notices
- Please note this event is virtual only.
- Please note that the recording link will be listed on this page when available.
Admission
Contact