Abstract:
|
Estimates of population characteristics such as domain means are often expected to follow qualitative assumptions. For example, average salary might be increasing in pay grade. Estimation and inference could be improved when these restrictions are incorporated along with sampling design of surveys, in comparison with the common methods that do not take account of these constraints. Particularly, domain estimation can be considerably improved when the correct shape restrictions are considered. Estimation of restricted domain parameters is accomplished through an adaptive domain estimation procedure. It consists on pooling the domains based on the assumed shape restrictions. However, even when shape restrictions can improve the precision of population estimates, it is important to use them appropriately. In fact, assuming incorrect restrictions could lead to biased estimators. We develop diagnostic methods that measure departures from the shape assumptions.
|