A continuous version of Liapunov's convexity theorem (Q1104440)

Let \((T,\Sigma,\mu_ 0)\) be a non-atomic probability measure space and \(\Phi\) the set of all positive finite measures \(\mu\) on T which are absolutely continuous with respect to \(\mu_ 0\). The metric in \(\Phi\) is induced by the norm \(\| \mu \|\) given by the variation of \(\mu\). A family \((A_{\alpha})_{\alpha \in [0,1]},\quad A_{\alpha}\in \Sigma,\) is called increasing if \(A_{\alpha}\subset A_{\beta}\) when \(\alpha\leq \beta\). For a compact metric space S with metric d, a map \(s\to (A^ s_{\alpha})\) is called continuous on S if for every \(s^ 0\in S\) and \(\epsilon >0\) there exists a \(\delta >0\) such that s, s' and s'' in \(B(s^ 0,\delta)\) implies \(\sup_{\alpha}\mu_ s(A_{\alpha}^{s'}\Delta A_{\alpha}^{s''})<\epsilon.\) In virtue of Liapunov's convexity theorem, it holds for a finite family of non-atomic measures \(\{\mu_ i\}_{1\leq i\leq n}\) that there exists an increasing family \((A_{\alpha})_{\alpha}\) such that \(\mu_ i(A_{\alpha})=\alpha \mu_ i(T),\quad i=1,2,...,n.\) But, in general, this is not true for an infinite family \((\mu_ s)_ s\). In this paper, the authors considered a problem of this line and obtained the following result: Let \(s\to \mu_ s\) be a continuous map from a compact metric space S into \(\Phi\). For every \(s\in S\) there exists an increasing family \((A^ s_{\alpha})_{\alpha}\) of measurable subsets of T satisfying \(\mu_ s(A^ s_{\alpha})=\alpha \mu_ s(T)\quad (\alpha \in [0,1])\) and such that the map \(s\to (A^ s_{\alpha})\) is continuous. But they also gave two examples where any constant selection cannot be guaranteed even under their setting.