|
Abstract:
|
Small area estimation (SAE) uses explicit or implicit statistical models to estimate characteristics for geographic areas or other domains where the available survey data are insufficient to produce acceptably reliable direct estimates. In practice, most SAE procedures are now based on small area models that explicitly account for the error in predicting the target characteristic given the auxiliary data. These SAE procedures can be classified as to whether the model is expressed at the area-level using direct survey estimates for the areas being modeled or at the unit-level, which is typically at the level of the individual survey response. In a 2016 paper, Hidiroglou and You examined the performance of some unit- and area-level SAE procedures for simulated samples from a known population, finding that unit-level estimators had a distinct advantage of area-level models. This paper expands their simulation to a broader set of circumstances and estimators in order to assess the generality of their findings.
|
