B1730
Title: Optimal estimation and sampling allocation in survey sampling under a general correlated superpopulation model
Authors: Ioulia Papageorgiou - Athens University of Economics and Business (Greece) [presenting]
Abstract: The aim is to contribute on the problem of sampling from autocorrelated populations. Assuming an autocorrelation among the population units we focus on deriving an efficient sample allocation and its resulting statistical inference about the population parameters. Both stages of sampling and inference incorporate the existing correlation. The optimal sample allocation is closely related with the type of correlation and therefore is a problem with no unique answer. The proposed methodology can cover any type of correlation function among the units. It is based on a continuous approximation of finite sums, is practically feasible and not computationally expensive. The result includes the sampling allocation that ensures optimal efficiency of the population parameters estimates and the expressions of the estimates and their mean square error. The gain in efficiency if the correlation type is taken into account is significant, and it is shown by a number of experiments with simulated data sets. An application in quality control with correlated measurements is also presented, aiming to exhibit the advantage of the proposed methodology over standard techniques that do not take into account the present correlation.
