Evaluation of Missing Data and Imputation Methods in Environmental Mixtures Analyses using Weighted Quantile Sum regression and Quantile g-computation Restricted; Files Only

Abstract

Background: Simultaneous exposure to multiple environmental chemicals poses challenges for traditional regression methods due to correlations among exposures and missing data issues. Weighted Quantile Sum (WQS) regression and Quantile g-computation (qgcomp) have emerged as valuable tools for analyzing mixtures of exposures. However, the impact of missing exposure data on these methods has received limited attention. 

Methods: We conducted a comprehensive simulation study to assess the performance of various imputation methods for missing exposure data in WQS and qgcomp analyses. We varied sample size, number of exposures subject to missingness, and percent of missingness to. Additionally, we applied these methods to real-world data from the Atlanta African American Maternal-Child Cohort to assess robustness of findings.

Results: Tree-based Imputation methods such as missForest and missRanger demonstrate best performance and robustness, particularly in scenarios with missing data due to limit of detection (LOD). Our results show that qgcomp generally outperforms WQS in bias, standard error, and 95% CI coverage. Incorporating the outcome in imputation had mixed effects on performance, indicating the need for careful consideration in applied analyses. In the application, estimates of the joint effect on birthweight remained robust across imputation methods, while estimates on gestational age at birth were more varied.

Conclusions: Our findings underscore the importance of considering missing data in exposure analysis. We also provide insights into the performance of imputation methods in WQS and qgcomp analyses and offer guidance for enhancing the reliability of findings.

Table of Contents

1      Introduction: 1

2      Methods: 2

2.1        Simulation Study: 2

2.2        Application to the Atlanta African American Maternal-Child Cohort: 3

3      Results: 3

3.1        Simulation Studies: 3

3.1.1     Joint Mixture Effect: 3

3.2        Analysis of the Atlanta African American Maternal-Child Cohort: 4

4      Discussion: 5

5      References: 8

6      Figures and Tables: 11

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Rollins School of Public Health
Department	Biostatistics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	M.S.P.H.
Submission	Master's Thesis
Language	English
Research Field	Health Sciences, Epidemiology Statistics Environmental Health
Palabra Clave	Imputation Chemical Mixtures Missing data Quantile g-computation Simulation Environmental epidemiology Weighted Quantile Sum regression
Committee Chair / Thesis Advisor	Howard Chang, Ph.D., Emory University
Committee Members	Stephanie Eick, Ph.D., MPH, Emory University

Última modificación

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	File download under embargo until 20 May 2026	2024-04-03 13:14:41 -0400	File download under embargo until 20 May 2026

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions
	File download under embargo until 20 May 2026	2024-04-03 13:14:48 -0400	File download under embargo until 20 May 2026
	File download under embargo until 20 May 2026	2024-04-03 13:14:58 -0400	File download under embargo until 20 May 2026
	File download under embargo until 20 May 2026	2024-04-03 13:15:11 -0400	File download under embargo until 20 May 2026
	File download under embargo until 20 May 2026	2024-04-03 13:15:25 -0400	File download under embargo until 20 May 2026
	File download under embargo until 20 May 2026	2024-04-03 13:15:43 -0400	File download under embargo until 20 May 2026