Simes Correction

Local clustering statistics calculated for a given section of a study area are not independent of one another, and hence their p-values in some instances (depending on one’s philosophical inclinations) should be corrected for multiple testing.

Lack of independence arises in the test statistics for neighboring areas. Such statistics are not independent because the z-scores for these areas enter into calculation of the Local Moran statistics for the neighboring areas.  Their p-values should be adjusted for this lack of independence.

SpaceStat adjusts the p-value using the Simes adjustment (Simes 1986), which is not as conservative as the Bonferroni correction. The Simes adjustment is calculated as in the equation below.

pi' = (n + 1 - a)pi

Here n is the number of p-values being considered (the number of neighbors + the central location), and a is the index (starting at 1) indicating the location in the sorted vector of the p-values for a location and its neighbors.