Logo image
Back
Bayesian spatial statistical methods for operationalizing and estimating the impact of racialized economic segregation in the US
Dissertation

Bayesian spatial statistical methods for operationalizing and estimating the impact of racialized economic segregation in the US

Yang Xu
Doctor of Philosophy (Ph.D.), Drexel University
Jun 2024
DOI:
https://doi.org/10.17918/00010639
pdf
Xu_Yang_202424.65 MB
PDF Embargoed Access, Embargo ends: 31 Aug 2026

Abstract

Bayesian statistical decision theory Segregation--Economic aspects
This dissertation consists of three papers that focus on the features and statistical assumptions of the Index of Concentration at the Extremes (ICE)--an operationalization of racialized economic segregation. Bayesian statistical frameworks, including parametric and semi-parametric methods, are developed to examine the relationship between the Index of Concentration at the Extremes and various health outcomes. We first start by developing a two-stage Bayesian statistical framework that provides a broad, flexible approach to studying the spatially varying association between premature mortality and racialized economic segregation while accounting for neighborhood-level latent health factors across US counties. We consider the spatial nature of racialized economic segregation and the health indicators used to construct latent health factors, which has been ignored in the existing studies. We show that the proposed two-stage framework reduces the dimensionality of spatially correlated data and highlights the importance of accounting for spatial autocorrelation in racialized economic segregation measures, in health equity focused settings. Then we propose a Bayesian semi-parametric model for estimating the varying relationship between racialized economic segregation and COVID-19 death, which allows the effect of racialized economic segregation to vary based on effect modifiers. The goal of method development is to tackle the non-linear relationship between racialized economic segregation and COVID-19 death as well as identify the effect modifiers that may cause the non-linearity, which has been revealed from previous research. We show that the proposed model could effectively capture the unknown function of the effect modifiers and improves model prediction through several simulation settings. Additionally, the proposed modeling methodology could have significant implications for public health agencies and state and local health departments when assessing the unequal burden of COVID-19 outcomes. Lastly, we propose reformulating the ICE metric using a Bayesian methodological framework, which enable us to quantify the uncertainty and the spatial correlation in the data, without losing the original interpretation of the ICE metric. Assuming a variety of residential segregation scenarios, we conduct simulation studies to evaluate the performance of our novel approach for reformulation of the ICE measure in comparison to other approaches often used when operationalizing residential segregation. Additionally, we present an application based on racialized economic segregation in Georgia, and using the proposed modeling approach, we examine whether racialized economic segregation has changed over two non-overlapping time points. We believe this work fills gaps in the existing literature by exploring various statistical assumptions of a racialized economic segregation measure found in ICE and its relationship to various health outcomes. Furthermore, the implications of the method development could provide public health researchers and state and local agencies with a high level of understanding of the ICE metric, influencing policy-making and practice, study design, and statistical modeling assumptions in residential segregation related research.

Metrics

79 Record Views

Details

Logo image