COMPSTAT 2022: Start Registration
View Submission - COMPSTAT2022
Title: A multi-spreading algorithm to account for spatial and strata heterogeneity Authors:  Maria Michela Dickson - University of Trento (Italy) [presenting]
Yves Tille - University of Neuchatel (Switzerland)
Giuseppe Espa - University of Trento (Italy)
Flavio Santi - University of Verona (Italy)
Diego Giuliani - University of Trento (Italy)
Abstract: Spatial designs producing spatially spread samples permit estimate precision to be improved when the variable of interest exhibits some form of spatial heterogeneity. However, if the variable of interests exhibits heterogeneity also with respect to some partition of the target population, a global spread of the sample over the reference space may not result in efficient estimates. On the other hand, the same problem may arise if sampled units are spatially spread only within each population stratum. We propose a sampling algorithm which considers both sources of heterogeneity and produces samples which are spatially spread both globally and within each stratum, according to a tuning parameter $\alpha\in[0,1]$, that globally spreads the sample when $\alpha=1$, and spreads each sub-sample independently when $\alpha=0$. Formal analysis and Monte Carlo simulations showed that the proposed algorithm can effectively exploit the trade-off between the global and within-stratum spread of the sample, and produce efficiency gains when the parameter is properly set.