|
Abstract:
|
Small area estimation techniques are used in survey sampling when sample sizes are too small to provide adequate precision for subpopulation estimates. These methods use implicit or explicit statistical models and are widely used in government agencies to plan action and make policy decisions. These model-based techniques require you to go beyond the design-based methods available in the SAS/STAT survey procedures; you need to use mixed effects modeling procedures to fit small area models. This workshop provides an overview of small area models and introduces SAS strategies to fit them. You will learn the two types of basic small area models, unit-level and area-level, which are illustrated with practical examples. Then, you will learn how to use the GLIMMIX and IML procedures to fit these models and obtain small area predictions and mean squared error for predictions with appropriate adjustments for the number of areas. Fully Bayesian approaches are commonly employed for small area estimation, too. You will also learn how to use the MCMC procedure to fit hierarchical Bayes models for small area estimation.
|
