This chapter developed a method for simulating spatially correlated beta-binomial data using
a transformation from a normal spatial field. The simulated spatial random fields were then
shown to follow the desired probability distributions, beta for the latent probability field
and binomial for the sample proportions. Some characteristics of the spatial relationships
in beta-binomial data were explored, which will be useful for comparing the beta-binomial
kriging model in later chapters. Spatial covariance parameter estimates were evaluated for
a large-scale simulation and the beta-binomial variogram estimator was shown to produce
accurate results. Finally the prediction accuracy of the beta-binomial kriging model was
assessed. Mean prediction error and mean squared prediction error are useful to compare
different cases of the beta-binomial kriging model, but they cannot say much about how
the model performs overall. In the next chapter these prediction summaries will be used to
compare alternative spatial models for binomial data.