ABSTRACT
In section 3.1.3 we saw that the generalized difference estimator (GDE)
tA = ∑
[Yi − Ai πi
] +
is a design-unbiased estimator of Y with A = (A1, . . . , Ai, . . . , AN )′ as a vector of known quantities and that it has optimal superpopulation model-based properties in case Ai = µi = Em(Yi), i = 1, . . . , N . But the µi ’s are usually unknown in practice.
