The Hausdorff metric of \(\sigma\)-fields and the value of information (Q2365741)

The Hausdorff metric \(\delta^ 1({\mathcal F},{\mathcal G})\) of two sub- \(\sigma\)-fields \({\mathcal F},{\mathcal G}\subseteq\Sigma\) is defined with the help of the measure \(\mu(F\triangle G)\) of the symmetric difference of the sets \(F\in{\mathcal F}\), \(G\in{\mathcal G}\). Another metric is \[ \delta^ 2({\mathcal F},{\mathcal G})= \sup_ f\| E(f|{\mathcal F})-E(f|{\mathcal G})\|_ 1, \text{ where } 0\leq f\leq 1. \] \textit{D. Landers} and \textit{L. Rogge} [Ann. Probab. 14, 724-730 (1986; Zbl 0597.60003); see also: Schriftenreihe des Fachbereichs Mathematik. Universität Duisburg Gesamthochschule, 56 (1984)] established inequalities between \(\delta^ 1\) and \(\delta^ 2\) which show that \(\delta^ 1\) and \(\delta^ 2\) are equivalent. The author introduced a new metric \(\delta^ 3({\mathcal F},{\mathcal G})\) using a metrization of stochastic convergence. \(\delta^ 3\) is shown to be equivalent to \(\delta^ 1\) and \(\delta^ 2\). As an application, the continuity of the value of information with respect to the Hausdorff metric of \(\sigma\)-fields is established.