The analysis of spatial data presumes the availability of geographic information associated with summary numbers of the phenomenon being analyzed such as incidents of measles in a geographic area. The relative position of the spatial units has a known location structure that is considered in the analysis – be areal units such as counties or geographic points of interest.
When geographical information is available, it is interesting to investigate if the distribution of the data displays a spatial pattern. The spatial distribution of the data may or may not follow a random pattern. Non-randomness may be due to the presence of either positive or negative spatial autocorrelation. Positive autocorrelation has clusters (e.g; cluster of counties) distributed in either a high-high (“hot-spot”) or low-low (“cold spot”) geographic layout. For example, a high concentration of crimes in neighboring counties characterizes a “hot-spot.” Negative autocorrelation has the units of analysis displaying values in significant contrast to the neighboring units characterizing a high-low or low-high spatial arrangement; for example, if measles outbreak occurs in isolated counties only but not in neighboring counties.