In unit-level small area models, the response variable corresponds to an individual element within a small area. Unit-level models serve a fundamental role in the field of small area estimation. Predictors based on unit-level models have been demonstrated to be more efficient than predictors based on area-level models, where the response variable is a direct estimator for an area. The seminal work on unit-level small area models is Battese, Harter, and Fuller (1988). This work uses unit-level small area models to predict crop areas at the county level. The model of Battese, Harter, and Fuller (1988) is linear and postulates normal distributions for the random terms. The assumptions of linearity and normality fail to hold in many practical situations. Therefore, research on the unit-level model has expanded in a variety of directions. Extensions of the Battese, Harter, and Fuller (1988) model include lognormal models, zero-inflated models, models with gamma response distributions, models for count data, and methods to incorporate an informative sample design. In this webinar, we discuss the use of the unit-level model for small area estimation, with emphasis on recent developments.
Contact: https:/