|
Activity Number:
|
176
- Contributed Poster Presentations: Section on Statistics and the Environment
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 31, 2017 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
|
Abstract #323781
|
|
|
Title:
|
Extending the Distributed Lag Model Framework to Evaluate Mixture Effects - a Nonparametric Approach
|
|
Author(s):
|
Ghalib Bello*
|
|
Companies:
|
Icahn School of Medicine at Mount Sinai
|
|
Keywords:
|
Distributed Lag Models ;
Nonparametric ;
Random Forests ;
Machine Learning ;
Multi-pollutant mixtures ;
Environmental health
|
|
Abstract:
|
Distributed Lag Models (DLMs) are used in environmental health studies wherein the effect of a pollutant is assumed to be distributed over time. Current values of an outcome are modeled as a linear function of prior (lagged) values of the exposure. While useful, the classic DLM formulation cannot adequately model complex mixture effects of multiple pollutants. Random Forest (RF) algorithm is a tree-based estimation technique that has shown strong performance in a wide spectrum of applications. Herein we demonstrate how RF can be used to extend the classical DLM to accommodate exposure to mixtures of pollutants, and to evaluate time-varying mixture effects. We tested the performance of tree-based DLMs using simulations of multi-exposure scenarios with complex, non-linear mixture effects. We examined multiple configurations arising from varying signal:noise ratio and correlations among mixture components. Tree-based DLM approach demonstrated the ability to capture these predefined functional relationships and showed robust performance across a variety of simulation settings. This study suggests the potential of tree-based DLM as a promising nonparametric alternative to classical DLMs
|
Authors who are presenting talks have a * after their name.
