An adaptive algorithm for estimating inclusion probabilities and performing the Horvitz-Thompson criterion in complex designs (Q638044)

The author considers empirical Horvitz-Thompson (eHT) estimators where the observations are weighted by weights reciprocal to some empirical estimates \(\hat p_j\) of the true first-order inclusion probabilities \(p_j\) of the sample units. The estimates \(\hat p_j\) are obtained by some replicated sampling scheme. An adaptive algorithm is proposed for determination of the necessary number of replications and calculation of eHT estimates. A two-phase Pareto sampling scheme and a spatial sampling in presence of spatial autocorrelations are considered as examples.