Order conditioned independence of real random variables (Q1061258): Difference between revisions

Denoting by \(<\) the usual partial ordering of finite measures on \({\mathbb{R}}\), the relation \(\mu_ 1<\mu_ 2\) is equivalent to the requirement that \(\mu_ 1\) and \(\mu_ 2\) occur as marginals of some measure \(\mu\) on \({\mathbb{R}}^ 2\) carried by the set F of points \((x_ 1,x_ 2)\) with \(x_ 1\leq x_ 2\). Moreover, the set M of all these measures \(\mu\) forms a compact convex set, possessing as typical extreme points measures carried by the graph of a bijection. The main aim of this paper is to find, in contrast, a concept of conditional independence by which a measure \(\mu\in M\) of ''maximal multiplicativity'' can be singled out in a unique (and explicit) way. To mention just one main result, which is of special importance for the Markov property of real point processes: a distribution on the set G of points \((x_ 1,x_ 2)\) with \(x_ 1<x_ 2\) can be rearranged in one and only one way such that (1) the marginals remain unchanged and (2) the restrictions to all rectangles contained in G are product measures.