Choquet order and simplices with applications in probabilistic models (Q1066414)

Let M be a convex set in a locally convex topological linear space E. An element x of M is represented by a probability measure p on M, i.e. it is the barycenter if \(\ell (x)=\int \ell dp\) for every \(\ell \in E^*\). One of the basic concerns in the theory of integral representation is to study those sets M where every element has a representation with a measure p living on the extreme points of M. In this connection the Choquet order plays a vital role. For measures p, q the Choquet order is defined by \[ p\prec q\quad iff\quad \int \psi dp\leq \int \psi dq \] for every convex continuous function \(\psi\). The classical books of Phelps (1966), Choquet (1969) and Alfsen and also more recent books of Asimow and Ellis (1980), Bourgin (1983) all had studied the Choquet ordering as a tool in the integration theory of various levels of generality. The present book of G. Winkler has somewhat different emphasis. The author starts studying Choquet order in its own right. Several equivalent formulations are in order: definitions by means of cones of functions, the dilation order and the description by means of conditional expectations which links the Choquet order and martingales. After presenting old and new results concerning Choquet order, the author proves some interesting inverse limit theorems of (compact or non- compact) simplices. These are then applied in concrete models from probability theory, e.g. in the construction of Dynkin's entrance boundary for Markov processes, and in the geometrical structure of sets of Gibbs states for statistical mechanics.