Fubini type products for densities and liftings (Q2385608)

In a former paper [Fundam. Math. 166, No. 3, 281--303 (2000; Zbl 0966.28001)], the authors studied densities and liftings with good section properties in the product of two probability spaces \(\Omega,\Sigma,\mu)\) and \((\Theta,T,\nu)\) such as \(\tau([ \varphi(E)]_\omega)=[\varphi(E)]_\omega\) for any \(E\) in their completed product \(\Sigma\widehat\otimes T\) and \(\omega\) outside a \(\mu\)-nullset, where \(\varphi\) is a primitive lifting for \(\mu \widehat\otimes\nu\). Such densities \(\tau\) for \(\nu\) are called \(\Theta\)-marginals, their existence was established in the former paper from Fund. Math. The main problem dealt with in this paper is, whether for given densities \(v\) for \(\mu\) and \(\tau\) for \(\nu\) it is possible to define a density or primitive lifting, at least, by means of the formula \((v \odot \tau)(E):=\{(\omega,\theta):\omega\in v(\{x\in\Omega:\theta\in\tau (E_x)\}) \}\) for \(E\in\Sigma\widehat\otimes T\). This product is called of `Fubini type', since it is an analogue of the classical Fubini-formula for measures. It traces back to the paper of \textit{N. D. Macheras} and \textit{W. Strauss} [Atti Sem. Mat. Fis. Univ. Modena L, 349--361 (2002; Zbl 1171.28302)]. It turns out, that the main problem is that of the measurability of the lifted set \((v\odot\tau)(E)\) for \(E\in\Sigma \widehat\otimes T\). By an application of hyperstonian spaces it can be seen, that the answer to this problem is in general to the negative, but if both \(v\) and \(\tau\) are marginals, the solution is to the positive. Even if the measurability is fullfilled, the hereditary properties of this product are bad. The product of two monotone liftings needs no longer to be monotone, and the product of two densities will in general be only a subdensity. The latter can be remedied by the introduction of the `upper hull' of the Fubini product. Moreover `subdensities' fall in the class of primitive liftings, and for this reason the above mentioned loss in structure foces to consider primitive liftings. In fact, there are more subtle investigations of this delicate measurability problem. The new methods centering about the Fubini product in this paper improve those of the former paper from Fund. Math, and provide a solution of one of the main problems of the Fund. Math. paper.