Chapter 6: Model-Assisted Estimation for Domains

In Chapter 6 of the textbook Practical Methods for Design and Analysis of Complex Surveys, estimation of population totals for population subgroups or domains is discussed by using design-based model-assisted estimation procedures. The generalized regression (GREG) estimator is used extensively. Different extensions are introduced where the alternatives of the assisting model cover linear fixed-effects models and linear mixed models involving domain-specific random effects in addition to the fixed effects. In addition to the GREG estimator, various synthetic estimators are introduced. Variance estimation of the estimated domain totals is discussed. The methods are demonstrated for the Occupational Health Care Survey data set.


In Training Key 190, the Horvitz-Thompson estimator and the GREG estimator of domain totals are compared with respect to the accuracy of estimation, reproducing the results of Example 6.1


In Training Key 205, the methods for domain estimation are examined further and the results of Example 6.2 are reproduced. An option is provided for more detailed examination of the accuracy of the estimators with different sample sizes.



NOTE: Instructions for the use of Training Keys are given in the Instructions section.