|
Activity Number:
|
50
- Advances in Spatial Statistics for Survey Methodology and Official Statistics
|
|
Type:
|
Topic-Contributed
|
|
Date/Time:
|
Sunday, August 8, 2021 : 3:30 PM to 5:20 PM
|
|
Sponsor:
|
Survey Research Methods Section
|
|
Abstract #317206
|
|
|
Title:
|
Hierarchical Bayesian Mixed Effect Models for Spatially Correlated Areal Count-Valued Data When Covariates Are Measured with Error
|
|
Author(s):
|
Saikat Nandy* and Scott Holan
|
|
Companies:
|
University of Missouri and University of Missouri
|
|
Keywords:
|
American Community Survey;
Bayesian hierarchical model;
Generalized Transformation Model;
Markov chain Monte Carlo;
Measurement error;
non-Gaussian data
|
|
Abstract:
|
We introduce hierarchical Bayesian models for predicting high-dimensional count-valued survey data which are non-Gaussian in nature. To incorporate dependence between covariates and geographic regions, we consider spatial mixed effect models that account for sampling error variance. In particular, we consider a Poisson data model with a latent Gaussian process model that assumes a structural measurement error model for the spatially correlated auxiliary variables. The proposed models are extremely high dimensional and employ the notion of Moran’s I basis functions to provide an effective approach to dimension reduction. To use a latent Gaussian process model structure, we utilize the hierarchical generalized transformation model mechanism proposed in Bradley et. al (2020) to transform the Bayesian model for a continuous response to a model that incorporates multiple response types. To demonstrate the utility and computational feasibility of our methodology, we provide the results of simulated examples and an application using a large dataset consisting of the US Census Bureau’s ACS 5-year estimates of the total count of population under the poverty threshold along census tracts.
|
Authors who are presenting talks have a * after their name.
