Chapter 13 Spatial Autocorrelation

In this chapter, I begin the discussion of spatial autocorrelation, both in general as well as focused on the special case of the Moran’s I statistic. I start off with some definitions, more specifically the notion of spatial randomness as the null hypothesis, and positive and negative spatial autocorrelation as the alternative hypotheses.

This is followed by an overview of the general concept of a spatial autocorrelation statistic, with a brief discussion of several specific implementations. Then attention shifts to arguably the best known such statistic, Moran’s I (Moran 1948). The formal structure, inference and interpretation of the statistic are covered, followed by its visualization through the so-called Moran scatter plot (Anselin 1996). The chapter closes with an outline of the implementation of the Moran scatter plot in GeoDa.

To illustrate these concepts, I use the Chicago Community Areas sample data set with socio-economic variables for the 77 community areas in the city of Chicago (from the American Community Survey).