2 EBLUPs and WFQ estimators
Yong You, J.N.K. Rao and Mike Hidiroglou
Previous | Next
Suppose that we have small areas with design-unbiased direct
estimators, of the area means The FH model refers to and associated area level auxiliary variables with Assuming independent sampling across areas,
the FH model may be written as a linear mixed model given by
(2.1)
where is the linking model, and is the sampling error with mean 0 and known
variance which is independent of the area specific
random effect Sampling is independent across areas and the are assumed to be independent and identically
distributed with mean 0 and variance
The best linear unbiased predictor (BLUP) of under the "true� model (2.1) is given by
(2.2)
where and is the optimal weighted least squared (WLS)
estimator of given by
(2.3)
see Rao (2003, page 116). The estimator depends on the unknown model variance and replacing in (2.2) by a suitable estimator we get the EBLUP:
(2.4)
where and is obtained from (2.3) by replacing by In this paper, we use the restricted maximum
likelihood (REML) estimator of assuming normality of and The weighted sum of the EBLUPs (2.4) does not necessarily agree
with the corresponding weighted direct estimator of the aggregate, where the are pre-specified weights such that is a design-consistent estimator of the aggregate
(total or mean). If the gap between and is large, it may indicate some model failure
that should be taken care of before proceeding to benchmarking, as noted by the
Associate Editor.
An estimator of the mean squared prediction error correct to second-order terms, under REML
estimation, is given by
(2.5)
where is the leading term of order and and are lower order terms of order accounting for the variability of and respectively (Rao 2003, page 128). We have
and
where is the estimator of The MSPE estimator (2.5) is nearly unbiased in
the sense that
WFQ obtained an EBLUP estimator, under the following augmented FH model:
(2.6)
where the random effects are independent with 0
and and is independent of The augmenting auxiliary variable is taken as WFQ showed that the EBLUP estimator of under the augmented model (2.6), is self-benchmarking in the sense of
satisfying
The EBLUP under the augmented model (2.6) is obtained
from (2.4) by changing to to and to The estimator of is similarly obtained from (2.5). Under the
true model (2.1), the MSPE of will be larger than the MSPE of but its estimator, will remain nearly unbiased under the true
model, as noted by the Associate Editor and a referee, because the true model
is a special case of the augmented model with
Previous | Next