Statistical Reliability of the Modified Areal Weighted by Control Zones Method to Spatially Downscale Individual Social Data

This study evaluates the modified areal weighting by control zones method (MAW-CZ) often involved in downscaling social data from a large spatial mesh, to a smaller mesh. This method has been extensively used in literature but the impossibil-ity, until recently, of accessing individual data makes it so that it has not been evaluated. In this study it is applied to two case studies, Toulouse and Grenoble-Alpes Metropoles, using the census INSEE data at the IRIS scale and the building islet or topographical reference units (RSU) scale. The study found that 27.2% of RSUs in the Toulouse metropolis and 21.9% in the Grenoble-Alpes metropolis are inhabited, with mean populations of 122 and 116 residents, and maximum populations of 2,429 and 6,451 residents, respectively in 2018. The chosen downscaling approach introduces small errors for small and medium-sizedRSUs. For example, 94%, 78%, and 72% of RSUs of <100, 101–255, and 256–500 inhabitants, respectively, are correctly classified by the modified areal weighting by control zones method in the Toulouse Metropole. However, there are significant differences for the most populated RSUs (the performance decreases to 60% for RSUs with more than 500 inhabitants), with this category having a representativeness of 8.4% and 7.2% of the total number of inhabited RSUs in the Toulouse and Grenoble-Alpes metropoles, respectively. The spatial distribution of the biased RSUs are nevertheless homogeneous throughout the two territories. These discrepancies are due to both the upscaling/downscaling methods used and the nature of the data (points in the upscaling and polygons in the downscaling)

Data mining · Data reliability · Geography · Physics · Statistics · Computer Science · Environmental Science · Mathematics · Soil Geostatistics and Mapping · Artificial Intelligence