|
Activity Number:
|
462
- Novel Spatial and Spatio-Temporal Models in Public Health
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
|
Abstract #312851
|
|
|
Title:
|
Treed Distributed Lag Nonlinear Models
|
|
Author(s):
|
Daniel Mork* and Ander Wilson
|
|
Companies:
|
Colorado State University and Colorado State University
|
|
Keywords:
|
regression trees;
distributed lag;
air pollution;
children’s health;
critical windows
|
|
Abstract:
|
The distributed lag non-linear models (DLNM) are used to estimate an exposure-time-response function when it is postulated the exposure effect is non-linear. Standard DLNM methods estimate an exposure-time-response surface parameterized with a bivariate basis expansion. However, basis functions such as penalized splines assume smoothness, which may be unrealistic when the exposure is associated with the outcome only in a specific time window. We propose a non-parametric framework for estimating the DLNM based on Bayesian additive regression trees. Our method operates using a set of regression trees that each assume piecewise constant relationships across the exposure-time space with boundary points estimated by dichotomous splits of the exposure and time dimensions. Each regression tree contributes a partial estimate and their sum allows for smoothness in the DLNM when necessary. In a simulation, we show advantages to our approach. We apply our non-parametric DLNM methods to estimate the association between maternal exposure to PM2.5 and birth weight in Colorado front-range births.
|
Authors who are presenting talks have a * after their name.
