\(\hat T_{GREG} = \hat T_{HT} + (T_x - \hat T_x)'B\). GREG is
where design-based and model-based thinking meet: the coefficient
B comes from a working model, but the estimator stays
design-consistent whether or not that model is right. A wrong
model costs EFFICIENCY, never validity, which is why GREG is the
standard production estimator in national statistics. Mirrors
morie.fn.genrgr.
Usage
morie_greg(y, x, weights, totals)
Arguments
- y
study variable.
- x
matrix of auxiliaries.
- weights
design weights.
- totals
known population totals of the auxiliaries.
Value
list: total, ht_total, correction, B, residual_totals,
design_consistent_regardless_of_model, n, p, method.
Examples
set.seed(1)
x <- cbind(1, runif(40, 0, 10))
morie_greg(rnorm(40), x, rep(20, 40),
c(800, 4000))$design_consistent_regardless_of_model
#> [1] TRUE
