Simple nonstandard proofs of Daniell-Kolmogorov-type theorems (Q1928254)

Suppose that \(X\) is a Polish space. For an infinite set \(I\), a projective family of probability measures over \(X\) and \(I\) is a family of probability measures \((\mu_J)\) on \(X^J\), where the indices \(J\) range over finite subsets of \(I\) and satisfy an appropriate coherency property, namely \(\mu_J=\mu_H\circ (\pi_{J}^{H})^{-1}\) whenever \(J\subseteq H\). The Daniell-Kolmogorov theorem for countable parameter sets states that whenever \(I\) is countable and \((\mu_J)\) is a projective family of probability measures over \(X\) and \(I\), then there is a probability space \((\Omega,\mathcal A, \mathrm{P})\) and an \(I\)-indexed family of \(X\)-valued random variables \(x_t\) such that, for each finite set \(J\subseteq I\), the measure \(A\mapsto \mathrm{P}((x_t)_{t\in J}^{-1}(A))\) coincides with the measure \(\mu_J\). This article gives a simple nonstandard proof of the Daniell-Kolmogorov theorem for countable parameter sets by considering the Loeb space attached to \(\mu_J\) for some hyperfinite sets \(J\) containing \(I\). The article also provides a simple nonstandard proof of the Daniell-Kolmogorov theorem for continuous projective families, where now \(I=\mathbb R^{\geq 0}\) and a suitable continuity requirement is imposed on the family \((\mu_J)\).