Implementations of two spatial-statistical plots added to standard R package

Two years ago, René Westerholt presented two diagrams at the GISRUK Conference 2024, the GIS conference for the UK and Ireland, which illustrate statistical inferences about spatial autocorrelation, particularly with regard to the Moran’s I statistic: the Moran drop plot and the Moran seismogram. Spatial autocorrelation is a statistical property that quantifies the strength and character of spatial structures. The plots presented have now been implemented in R and integrated into the ‘spdep’ package, which is one of the standard R packages for spatial analysis. The plots are not yet available via CRAN (this should be the case shortly), but they can already be used if spdep is installed directly from GitHub. The corresponding functions are called ‘moran.plot.drop’ and ‘moran.plot.seismogram’. The drop plot visualises statistical p-values using lines that ‘drop down’ in the plot from the observed autocorrelations to the (hypothetical) coordinates of the respective critical values for assessing statistical significance. The longer these lines are, the less likely it would be to observe a locally observed spatial structure by chance. The second plot, called the seismogram, connects these coordinates of the critical values with lines across all displayed observations to highlight potential patterns or outliers and thus potentially noteworthy local spatial configurations across the entire range of attribute values. The stronger the spikes in the seismogram, the more exceptional is the spatial arrangement of polygons, points, or lines on which the analysed attribute values were collected.
Westerholt, R. (2024): Extending the Moran scatterplot by indications of critical values and p-values: introducing the Moran seismogram and the drop plot. 32nd Annual GIS Research UK Conference (GISRUK), Leeds, UK. DOI: 10.5281/zenodo.10897792
