Dissertation
Statistical approaches for examining exposure to multiple environment features
Doctor of Philosophy (Ph.D.), Drexel University
Aug 2024
DOI:
https://doi.org/10.17918/00010764
Abstract
It can be difficult to estimate effects of the built environment due to it's multi-faceted interrelated features that may operate on different spatial scales. In this dissertation we develop three approaches to characterize one aspect of the built environment, the food environment (FE), and examine it's effect on measures of health. We first propose a latent transition analysis (LTA) model to classify the FE and its evolution over time. The model uses zero-inflated Poisson distributions to model count data with an excess of zeros. Using simulations, we compare it's performance to more commonly used LTA models that model observed variables using Binary or Multinomial distributions. Secondly, we develop a bias-adjusted 3-step BCH-GEE approach to estimate the effect of latent classes on multi-level longitudinal outcomes. In simulation studies we investigate how it compares to a standard 3-step GEE approach and how the correlation structure used impacts it's performance. Lastly, we propose LASSO methods, built upon the fused and clustered LASSO, to address questions such as spatial scale and feature grouping that arise when estimating the effects of the FE on health outcomes. We apply these methods to characterize the unhealthy food environment surrounding California public schools, as well as estimate the association between child obesity, and their schools' food environment.
Metrics
5 File views/ downloads
46 Record Views
Details
- Title
- Statistical approaches for examining exposure to multiple environment features
- Creators
- Kelsey Alexovitz
- Contributors
- Brisa Sanchez (Advisor)
- Awarding Institution
- Drexel University
- Degree Awarded
- Doctor of Philosophy (Ph.D.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- xi, 105 pages
- Resource Type
- Dissertation
- Language
- English
- Academic Unit
- Dana and David Dornsife School of Public Health; Epidemiology and Biostatistics; Drexel University
- Other Identifier
- 991021906712404721