Extreme probability submeasures on 3 points (Q1890602): Difference between revisions

A probability submeasure \(\mu\) on \((X, {\mathcal P} (x))\) is a function on \({\mathcal P} (X)\), satisfying \hskip10mm 1. \(\mu (\emptyset)= 0\leq \mu(A)\leq 1= \mu(X)\); \hskip10mm 2. \(\mu(A)\leq \mu(B)\) whenever \(A\subseteq B\); \hskip10mm 3. \(\mu(A\cup B)\leq \mu(A)+ \mu(B)\). If \(X\) is a finite set, the collection of probability submeasures \({\mathcal S}\) on \((X, {\mathcal P} (X))\) forms a compact, convex set. In this paper, the authors solve the problem (due to J. Roberts) to determine the extreme points when \(X= \{1, 2, 3\}\). There are twelve extreme submeasures \(\varphi_i\), such that every submeasure \(\mu\) on \(X\) is a convex combination of \(\varphi_i\)'s. Few lemmas which apply in the general case are demonstrated.