Chapter 16 LISA and Local Moran

This chapter is the first in a series where attention shifts to the detection of the location of clusters and spatial outliers. These are collections of observations that are surrounded by neighbors that are more similar to them or more different from them than would be expected under spatial randomness. The general principle by which this is approached is based on the concept of Local Indicators of Spatial Association, or LISA (Anselin 1995).

I begin by outlining the basic idea underlying the concept of LISA. This is followed by an in-depth coverage of its most common application, in the form of the Local Moran statistic. This statistic becomes a powerful tool to detect hot spots, cold spots, as well as spatial outliers when combined with the classification of spatial autocorrelation in the Moran scatter plot. The ultimate result is a local cluster map. An extensive discussion of its properties and interpretation is provided, with special attention to the notion of significance.

To illustrate these methods, I will employ the Oaxaca Development sample data set.