Uniform measures on linear spaces (Q1098958): Difference between revisions

Let E be a locally convex Hausdorff space and C(E,E') be the cylindrical \(\sigma\)-algebra generated by the topological dual E'. Let \(\mu\) be a probability measure on \(C(E,E')\) and \(\tau_{\mu}\) be the topology of convergence in measure on E' semi-metrized by \[ d(x',y')=\int_{E}| x'(x)-y'(x)| /(1+| x'(x)-y'(x)|)d\mu (x). \] For each A in C(E,E') with \(\mu (A)>0\), we put \(\mu_ A(B)=\mu (A\cap B)/\mu (A)\) for all B in \(C(E,E').\) We say \(\mu\) is uniform if \(\tau_{\mu_ A}\) is finer than \(\tau_{\mu}\) whenever \(\mu (A)>0.\) The notion of uniformness has been introduced by \textit{R. M. Dudley} [Z. Wahrscheinlichkeitstheor. Verw. Geb. 6, 129-132 (1966; Zbl 0303.28012)]. We prove that \(\mu\) is uniform if and only if for every \(x_ n'\) in E', there exists a subsequence \(x'_{n_ j}\) such that \(x'_{n_ j}\to 0\mu\)-almost everywhere (Theorem 1). We also prove that \(\mu\) is uniform if and only if for every probability measure \(\nu\) which is not singular with \(\mu\), \(\tau_{\nu}\) is finer than \(\tau_{\mu}\). By the above results we obtain that if \(\mu\) and \(\nu\) are stable measures, or more generally, semi-stable measures, then \(\tau_{\mu}\) and \(\tau_{\nu}\) are equivalent, or, \(\mu\) and \(\nu\) are singular. This extends a result of \textit{J. Zinn} [Stud. Math. 54, 245-257 (1976; Zbl 0339.60003)].