Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 458 - Models for Spatial and Environmental Data
Type: Contributed
Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
Sponsor: Section on Statistics and the Environment
Abstract #312329
Title: Quantile Regression for Exposure Data with Repeated Measures in the Presence of Non-Detects
Author(s): I-Chen Chen* and Stephen J. Bertke and Brian D. Curwin
Companies: National Institute for Occupational Safety and Health and National Institute for Occupational Safety and Health and National Institute for Occupational Safety and Health
Keywords: quantile regression models; left censoring; right skewness; limit of detection; repeated measures; environmental exposure
Abstract:

Exposure outcome data with repeated measures from occupational studies are frequently right-skewed and left-censored. The data are generally transformed using natural logarithms and assumed to follow a log-normal distribution. However, modeling the conditional mean may not be ideal if the transformed data don’t follow a known distribution. Even when distributional assumptions are met, it can lead to bias and less precision for finite-sample setting. Left censoring occurs when measures are below limit of detection (LOD). Substitution approach replacing a value for measures less than the LOD is regularly used, but the regression parameter estimation can be biased for large censoring. Therefore, maximum likelihood approach has been advocated to reduce this bias. Unfortunately, this method works poorly for highly skewed data, even if the data distribution is correctly specified. We propose a quantile regression model that has advantages for skewed data, does not require any error distributions, and can provide a complete illustration of the entire conditional distribution of an outcome variable. Existing and proposed methods are compared in a simulation study and application example.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program