ABSTRACT
This chapter presents methods for exploring the mean structure and covariance structure of spatial data, which help to specify spatial linear models likely to be useful for a particular spatial dataset. Such methods include stem-and-leaf plots and maps of data sites, although these pieces of information only consider the observations and data sites separately. Other methods presented for exploring geostatistical data include three-dimensional scatterplots, choropleth maps, and bubbleplots for exploring large-scale variation; nearest-neighbor scatterplots for detecting outliers; plots of subregion-specific variances to explore spatial homogeneity of variance; and the sample autocovariance function and sample semivariogram for exploring small-scale variation. Methods presented for exploring areal data include neighbor graphs, autocorrelation statistics such as Moran's I, correlograms, and local indicators of spatial autocorrelation (LISAs).
