Understanding and Modeling the Uncertainty in the American Community Survey: Detecting the Spatial Correlation in Uncertainty at the County Level with Conditional Autoregressive Models Open Access

Tang, Shiwei (2017)

Permanent URL: https://etd.library.emory.edu/concern/etds/9019s328x?locale=en


Starting in 2010, the US Census Bureau replaced the Long Form with the American Community Survey (ACS), a rolling sample with annual and five- year data summaries. Because of the sampling design, the ACS reports the uncertainty of estimation via a margin of errors, which can be used to model the precision of results. Recent research illustrates latent spatial correlation in these margins of errors at different level of aggregation, and conditional autoregressive (CAR) models are popular way to incorporate such spatial correlations. In this thesis, we use spatial generalized linear mixed models (GLMMs), based on Poisson regression and the CAR model, to incorporate uncertainty in ACS-based population counts and covariates in small area estimates of disease risk. Both Markov chain Monte Carlo (MCMC) and Integrated Nested Laplace Approximation (INLA) are used to implement GLMMs, and we find both methods provide similar results. However, INLA is much more efficient in computation while MCMC provides slightly better interval coverages for fixed parameters.

Table of Contents

1 Introduction

1.1 Modeling

1.2 Data Sources

2 Methods

2.1 Simulation model

2.2 Fitted model

2.3 Comparsion Techniques

2.4 Data Summaries

3 Results

4 Discussion

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.
Subfield / Discipline
  • English
Research Field
Committee Chair / Thesis Advisor
Committee Members
Partnering Agencies
Last modified

Primary PDF

Supplemental Files