While these formulas do not represent the unequal sample size case, they illustrate
an important point about the relative size of the nugget effect of the binomial sample
proportions. The nugget effect ⌧ 2
Z changes by a factor of 1
n, as the sample size grows at
each spatial location the nugget effect diminishes quickly. So, the adjustment provided by
beta-binomial kriging is most important when there are relatively few observations at each
sample location.
At this point, reconsider Figure 3.4 and the variances of the binomial sample proportions.
The overall variance of a spatial data set can be decomposed into two components: the
nugget effect and the spatial sill.