Small-Area Estimation for Iowa Seat-Belt Use Survey in 2017-2021

Abstract

This study addresses the challenges in small-area estimation for the Iowa Seat-Belt-Use survey data. The data consists of multivariate counts and proportions that are not normally distributed, with many observed proportions equaling 1. Additionally, the data is longitudinal, with the same sample of road segments revisited over 5 years. To address these challenges, we propose a multi-level small area model that integrates data from multiple time points and respects the distribution of the observed data. We explore multiple distributions for the observed proportions and integrate the Poisson model for counts, incorporating multivariate random effects to reflect the bivariate data structure.

Keywords: Bayesian, Survey, Seat-belt, INLA, GLMM, Statistics