Finite additivity and closest approximations (Q1898962)

This paper addresses nearest point problems for certain sets of finitely additive measures. While certain subsets of the space of bounded finitely additive real valued measures defined on a field of sets are known to have the unique nearest point property, the results of the present paper apply more broadly to sets of unbounded measures and to situations, where nearest points may not be unique. For a fixed field \(\mathcal F\) and a finitely additive real valued measure \(\mu\) on \(\mathcal F\), let \(A(\mu)\) represent the set of finitely additive real valued measures \(\nu\) on \(\mathcal F\) for which \(\mu- \nu\) is bounded. Suppose \(M\subseteq A(\mu)\) is closed under (i) supremum and infimum of pairs of measures, (ii) sums of pairs restricted to complementary sets (measures of the type \(E\mapsto \nu(E\cap F)+ \eta(E\backslash F)\) for a fixed \(F\in {\mathcal F}\)), and (iii) setwise supremum and infimum of monotone sequences when these are finite. Then there is an element \(\nu\) of \(M\) nearest to \(\mu\) in the sense of total variation. Furthermore, given any two finitely additive real valued measures \(\mu\) and \(\nu\) on \(\mathcal F\) for which \(M\subseteq A(\mu)\cap A(\nu)\), and elements \(\mu^0\) and \(\nu^0\) of \(M\) nearest to \(\mu\) and \(\nu\), respectively, \(\sup\{\mu^0, \nu^0\}\) is an element of \(M\) nearest to \(\sup\{\mu, \nu\}\). Conversely, any element of \(M\) nearest to \(\sup\{\mu, \nu\}\) can be expressed as the supremum of two members of \(M\) nearest to \(\mu\) and \(\nu\), respectively. These results are obtained using notions of refinement integration and related techniques such as differential equivalence.