Each of the panels above features a histogram of neighbor counts for each weight matrix. For example, for Rook Contiguity (w_rook), five is the most frequent neighborhood size. This is in sharp contrast to the Kernel matrix (w_kern), which has only six counties with a neighborhood size of five. Visual inspection of the neighborhood size distribution demonstrates the substantive variation in the definitions employed by each weight matrix.

With the neighborhoods defined, LISAs can be evaluated. There are two primary measures used to establish spatial clustering for each weight matrix/year combination. Morans's I measures the global spatial autocorrelation in an attribute y over a given neighborhood defined by w_ij. The local version is implemented as follows: