Measures with prescribed marginals, extreme points and measure preserving transformations (Q809211)

Let (X,\({\mathcal A},\lambda)\), (Y,\({\mathcal B},\nu)\) be two probability spaces, let M(\(\lambda\),\(\nu\)) be the set of all probability measures \(\mu\) on the product space \(X\times Y\) such that the marginals of \(\mu\) are \(\lambda\),\(\nu\) respectively. The authors give a method for constructing certain extreme points of the convex set M(\(\lambda\),\(\nu\)); the method is based on the existence of a transformation T: \(X\to Y\) which transforms \(\lambda\) to \(\nu\). Let G be the graph of T and define \(\mu_ T(E)=\lambda (P_ 1(E\cap G)),\) where \(P_ 1\) is the projection from \(X\times Y\) to X; it is shown that \(\mu_ T\) is an extreme point of M(\(\lambda\),\(\nu\)) in the sense that if \(\mu_ T=(1/2)(\zeta +\eta)\) then \(\zeta =\eta =\mu_ T.\) References to previous work are given and some remarks pointing out the existence or non-existence of other types of extreme points are made.