scientific article; zbMATH DE number 1021815

Contents, measures, outer measures, capacities (28A12) Measures and integrals in product spaces (28A35) Convex sets in topological linear spaces; Choquet theory (46A55)

Review text

Let \(\Omega\) be a set and let \({\mathcal A}\) be a \(\sigma\)-algebra of subsets of \(\Omega\). For \(A\in{\mathcal A}\) and \(\mu\in\text{ca}_+({\mathcal A})\) define \(\mu_A\in \text{ca}_+({\mathcal A})\) by \(\mu_A(C)= \mu(A\cap C)\). Let \({\mathcal B}\) be another \(\sigma\)-algebra of subsets of \(\Omega\) with \({\mathcal B}\subset {\mathcal A}\). The author says that \({\mathcal A}\) is \({\mathcal B}\)-approximable with respect to \(\mu\) if for every \(A\in{\mathcal A}\) there exists \(B\in{\mathcal B}\) with \(\mu(A\triangle B)= 0\). Let, moreover, \(\nu\in \text{ca}_+({\mathcal A})\) and denote by \(S_{\mu,\nu}\) the convex set of all \(\lambda\in\text{ca}_+({\mathcal A})\) with \(\lambda\leq\mu\) and \(\lambda|{\mathcal B}= \nu|{\mathcal B}\). It is a consequence of the Krein-Milman theorem that \(S_{\mu,\nu}\) has extreme points. The author characterizes them as follows. For \(\lambda\in S_{\mu,\nu}\) we have \(\lambda\in \text{extr }S_{\mu, \nu}\) if and only if \({\mathcal A}\) is \({\mathcal B}\)-approximable with respect to \(\lambda_{A_0}\), where \(A_0= \{0< d\lambda/d\mu< 1\}\) (the first part of Theorem 1). The results is much in the spirit of a classical theorem due to R. G. Douglas and J. Lindenstrauss, and the proof is essentially the same. As an application, it is shown that for \(\mu\in \text{ca}_+({\mathcal A})\) the following two conditions are equivalent: (1) There exists no nonzero \(\lambda\in \text{ca}_+({\mathcal A})\) such that \({\mathcal A}\) is \({\mathcal B}\)-approximable with respect to \(\lambda\) and \(\lambda\leq\mu\); (2) Given \(k>1\) and \(A\in{\mathcal A}\) there exists \(A_k\in{\mathcal A}\) with \(A_k\subset A\) and \(\mu_A|{\mathcal B}= k\mu_{A_k}|{\mathcal B}\). This is a special case of a result of H. Luschgy who also proved that each \(\varphi\in \text{ca}_+({\mathcal A})\) has a largest minorant \(\mu\in \text{ca}_+({\mathcal A})\) which satisfies (1) [\textit{H. Luschgy}, Proc. Am. Math. Soc. 96, 121-125 (1986; Zbl 0605.28003)]. The author is also concerned with the identification of that minorant in the case where \(\varphi\) is the direct product of two measures and \({\mathcal B}\) is a product \(\sigma\)-algebra (Theorem 3). \{Reviewer's remarks: (1) The second part of Theorem 1 can be proved as follows. With the notation of the author, set \(C=\{d\lambda/d\mu= 1\}\) and \(D= \Omega\backslash C\), and assume \(\lambda(C)= \varrho(C)\). Since \(\lambda_C= \mu_C\geq \varrho_C\), we have \(\lambda_C= \varrho_C\). We also have \(\lambda|{\mathcal B}= \varrho|{\mathcal B}\), and so \(\lambda_D|{\mathcal B}= \varrho_D|{\mathcal B}\). Moreover, \(\varrho_D\ll \lambda_D\) and \({\mathcal A}\) is \({\mathcal B}\)-approximable with respect to \(\lambda_D\). It follows that \(\lambda_D= \varrho_D\). Hence \(\lambda= \varrho\). (2) In the example discussed on pp. 239-241 one can take for \(\Omega\) a two-element set. The assertion then becomes an exercise from analytic geometry. (3) The assertion of Example 1 is also contained in a paper by the reviewer [Manuscr. Math. 86, No. 3, 349-365 (1995; Zbl 0826.28004), Theorem 1(a)]. (4) In the proof of Theorem 3 it is not clear why one can choose \(A_0\) in \({\mathcal B}\otimes{\mathcal B}'\) so that \({\mathcal A}\otimes{\mathcal A}'\) is \({\mathcal B}\otimes{\mathcal B}'\)-approximable with respect to \((\mu_1\otimes \mu_2)_{A_0}\)\}.

Recommendations

Cited in

(11)

This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4339656)