A simpler way of obtaining a `model for clustering' is the following.
For each point 
define
Then 
is just twice the neighbour pair statistic
s(x) used in defining the Strauss process and the triplets process. But
instead of adding the 
we put an upper bound on the influence of
any single point. Let c ;SPMgt; 0 be an arbitrary constant, and define
Define t(x) to be the 2-dimensional vector (n(x), u(x)). Then the exponential family having canonical statistic t(x) and unnormalized densities (1.3) is called here the saturation process.
