It is based on calculation of a test statistic, z, that has an asymptotic
standard normal distribution under the null hypothesis. It is one-sided (in
the way that the alternative is aggregation, not just "non-randomness"), thus
with a confidence level of 95
1.64. When all the sampling units contain the same total number of
individuals, n, the test statistic is calculated from:
z = (n(N - 1)I - Nn)/(2Nn(n - 1))^(1/2)
where N is the number of sampling units, and I, Fisher's index of aggregation
for incidence data.