B1373
Title: Functional and structural measurement error models with global-local priors for random effects in small area estimation
Authors: Xueying Tang - University of Arizona (United States) [presenting]
Jairo Fuquene - UC Davis (United States)
Abstract: Small area estimation (SAE) plays an important role in producing economic and public health indicators in statistical offices of developing countries. A common method of SAE utilizes the Fay-Herriot model to estimate small area means based on the direct estimates from surveys with the help of auxiliary covariates from administrative records and/or population census. Due to the long intercensal periods and the incompleteness of administrative records in developing countries, finding covariates for small area estimation is often challenging, and estimates from another survey are used as an alternative. This calls for models that account for the measurement errors of the auxiliary variables. Existing measurement error models often assume homogeneous variance for random effects across small areas. However, it has been shown that this assumption is often invalid and that taking into account the heterogeneity of random effects could increase the accuracy of SAE. We use global-local priors for the random effects in the functional and structural measurement error models to accommodate heterogeneous random effect variance. We theoretically examine the behavior of the posterior mean estimator under the proposed models. MCMC algorithms are designed to obtain posterior samples of the model parameters. We demonstrate the performance of the proposed models through a simulation study and a case study of estimating public health indicators in the municipalities of Colombia.
