Bayesian Models for the Aggregation of Public Opinion Survey Data
Abstract
Public opinion polling data has unique features that complicate statistical inference. Polling
data is often non-representative of the population it aims to estimate. It is also common for individual polls to have a large enough measurement error to make estimating the population mean difficult. Previous research has developed various methods for analyzing individual polls and other methods for polling aggregation. This dissertation will focus on three distinct but related models. The first is a model to predict American presidential election data. The second is a combination of Bayesian model averaging and multiple imputation to create regressions based on a panel survey about terrorism policy preferences. The last is a Bayesian model based on combining two distinct panel surveys about terrorism with different sampling frames and measurement schedules. These three models apply to other problems and address common problems in analyzing public opinion.
Citation
Alexander, Brittany Marie (2022). Bayesian Models for the Aggregation of Public Opinion Survey Data. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /197102.