Abstract:

Spatial data have become increasingly relevant to economic surveys, reflecting recognition that economic outcomes vary substantially across geographic space and require spatially disaggregated analysis. However, in this context, incorporation of spatial distribution as part of the sample design remains limited. This work develops a framework for multivariate spatial stratified sampling using SKATER and REDCAP algorithms for stratification combined with adapted Neyman allocation for sample allocation with multiple stratifiers. Through an application to the South African Spatial Economic Activity Data, we demonstrate that spatial stratified sampling is able to better balance precision and spatial representation in comparison to traditional stratified sampling and spatially balanced sampling methods.

About the presenter:

Georgi Borros is a Research Manager at SALDRU's Survey & Data Hub, where she works on survey design and implementation. She is currently completing her PhD in Statistical Sciences at the University of Cape Town, focusing on survey methodology and sampling. The research presented in this talk forms part of her PhD thesis, Optimal Stratified Sampling: Novel Designs for Application in South African Surveys.