Implementing a weighted measure of multivariate spatial autocorrelation

Bavaud (2024) builds on and significantly broadens earlier work on measuring spatial autocorrelation, extending to multivariate settings and in particular regional weights. While measurement of spatial autocorrelation in multivariate data has been approached previously, the addition of regional weights is a major advance, as regions often differ in their contributions to global measures. Thus far, the proposed implementation described here involves dense matrices, as does much multivariate analysis. It is shown that the implementation largely reproduces the results presented in Bavaud (2024).