The Swiss Federal Office of Statistics uses a coordinated sampling system developed in Qualité (2009), and that is an extension of the method proposed in Brewer et al. (1972). This system allows for the selection of negatively or positively correlated samples, panels and rotating panels in a dynamic population. Each transversal sample selected through this system stems from a Poisson sampling design. This procedure, with its inherent size-variability, calls for the use of calibrated estimators and updated planification methods of target sample sizes within domains, to replace allocation optimization techniques that were used for stratified fixed-size designs. One particular aspect, introduced with these Poisson designs, and that did not exist with fixed-size simple random or stratified sampling designs that were previously used, is the risk of selecting a sample that is well below the expected size in some domains. A risk that is also present when non-response is modeled as a second-phase Bernoulli sampling within domains, and taken into account at the planification stage.
Another problem for which we give our best-yet-found solution, is the unit-level selection dependence required in some surveys (e.g. surveys where at most one selection may occur in any household). A dependence that is unobtainable through a simple use of our coordinated sampling system. In order to satisfy these requirements, we propose to use coordinated selection as one phase in a multi-stage sampling design. The computation of correct (conditional-)inclusion probabilities at each phase is however non-trivial in the general case.
Brewer, K., Early, L., and Joyce, S. (1972). Selecting several samples from a single population. Australian Journal of Statistics, 3:231–239.
Qualité, L. (2009). Unequal probability sampling and repeated surveys. Doctoral Thesis. University of Neuchâtel, Switzerland.
Keywords: Coordinated sampling; Sample allocation
Biography: Mr Qualité works part-time as a scientific collaborator for the university of Neuchâtel and part-time for the methodology section of the Swiss Federal Office of Statistics