image

where Xm,c is the total payroll in the nonsampled portion of state m. Finally, as an analogue of the modified direct estimator,

image

where

image

Final Considerations and Discussion

Gershunskaya concluded her review of small-area estimation methods with a set of important considerations:

• It is important to plan for estimation for domains of interest at the design stage to ensure that one has direct estimates of some reliability to start off.

• Finding a set of good auxiliary variables is crucial for success in small-area modeling.

• Small-area estimation methods are based on assumptions, and therefore evaluation of the resulting estimates is vital.

• Using a statistical model supports a systematic approach for a given problem: (a) the need for explicitly stated assumptions, (b) the need for model selection and checking, and (c) the production of measures of uncertainty.

• It is important to account for the sample design (unequal probabilities of selection and clustering effects) in the model formulation and fitting.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement