Activity Number:
|
337
- Environmental Epidemiology and Analysis of Large Database
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 9, 2022 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Statistics in Epidemiology
|
Abstract #322837
|
|
Title:
|
Nonparametric Screening and Selection in Presence of Dependence Among Predictors
|
Author(s):
|
Shanta Ghosh* and Sanjib Basu
|
Companies:
|
University of Illinois at Chicago and Biostatistics, University of Illinois Chicago
|
Keywords:
|
Variable selection;
Environmental exposures;
Collinearity;
High dimensional;
Distance Correlation
|
Abstract:
|
Variable selection under multicollinearity is well studied but remains challenging. Our motivating application arises in environmental epidemiology where individuals are exposed simultaneously to a multitude of pollutants in the environmental mixture that potentially interact and present a health risk. The pollutant measures are often highly correlated at levels that are generally not seen in other areas of science. We develop a model-free screening and selection method using distance correlation, which is a non-parametric measure of dependence in arbitrary dimensions. We compare performance with existing methods under linear and nonlinear data-generating models. We apply the proposed method to environmental mixtures data in NHANES involving many strongly correlated persistent organic pollutants.
|
Authors who are presenting talks have a * after their name.