Activity Number:
|
322
- Analyses in Ecology, Epidemiology, and Environmental Policy
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 11, 2021 : 3:30 PM to 5:20 PM
|
Sponsor:
|
WNAR
|
Abstract #318376
|
|
Title:
|
Predicting a Spatiotemporal Exposure Surface with Penalized Regression
|
Author(s):
|
Nathan Ryder* and Joshua Keller
|
Companies:
|
Colorado State University and Colorado State University
|
Keywords:
|
particulate matter;
air pollution;
convex optimization;
spatial statistics;
downscaler models;
epidemiology
|
Abstract:
|
Acute and chronic exposures to fine particulate matter (PM2.5) have been linked to premature mortality from several health effects. PM2.5 is measured by hundreds of monitors across the US, and must be predicted across space and time for epidemiological analysis in large cohorts. We propose a penalized regression model to predict a spatiotemporal surface with reduced computational cost. The model penalizes to reduce overfitting and to encourage smoothness in time. A model coefficient is estimated for each predictor at each time point, allowing us to predict a response surface for any time in the domain. To predict spatially, the model can incorporate spatial splines (like thin-plate regression splines) and other location-specific spatial and/or temporal predictors. We demonstrate the method by predicting daily PM2.5 concentrations across the contiguous United States for a five-year period. Our model is compared with existing methods by predictive accuracy and computation time.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2021 program
|