Small area (or small domain) estimation methods are a set of techniques allowing the estimation of parameters of interest for domains where the direct estimators (e.g., HT or GREG; see the theme module “Weighting and Estimation – Main Module” and the method module “Weighting and Estimation – Generalised Regression Estimator”, respectively) cannot be considered reliable enough, i.e., their variance is too high to be released. National Statistical Office surveys are usually planned at a higher level, hence, whenever more detailed information is required, the sample size may be not large enough to guarantee release of direct estimates and in some cases, smaller domains may happen to be without sample units. Small area methods increase the reliability of estimation by “borrowing strength” from a set of areas in a larger domain for which the direct estimator is reliable. This means that information from other areas is used and/or additional information from different sources is exploited (see the theme module “Weighting and Estimation – Small Area Estimation”).
The area level EBLUP, which is described in this module, is a linear combination of the area (domain) direct estimator and a predicted component based on a linear mixed model. The model relates the parameter of interest to known auxiliary variables for each of the domains that constitute the partition of the whole population. An effect to account for (within) domain homogeneity is included in the model.
To read the entire document, please access the pdf file (link under "Related Documents" on the right-hand-side of this page).
Your feedback is appreciated. Please send your remarks, suggestions for improvement, etc. to email@example.com.