A bipolar theorem only using independent random variables
The author continues his study from [\textit{F.~Delbaen}, Finance Stoch. 25, No.~3, 597--614 (2021; Zbl 1470.91108)] of probability spaces $(\Omega,\mathcal F,\mathbb P)$ such that $\mathcal F$ is atomless conditionally to a sub-$\sigma$-algebra $\mathcal G\subseteq \mathcal F$, which means that, for every $A\in\mathcal F$, there is $B\in\mathcal F$ with $B\subseteq A$ and $0<\mathbb P(B\mid \mathcal G) < \mathbb P(A\mid \mathcal G)$ a.s. on $\{\mathbb P(A\mid \mathcal G)>0\}$. This property is characterized by the existence of a random variable on $(\Omega,\mathcal F,\mathbb P)$ which is independent of $\mathcal G$ and uniformly distributed in $[0,1]$.\N\NIn this setting, the author determines the extreme points of the unit ball $\mathcal Z$ of the Banach space \N\[\NE=\{\eta\in L^\infty(\Omega,\mathcal F,\mathbb P): \mathbb E(\eta\mid \mathcal G)=0\}\N\]\Nas those elements of $E$ satisfying $\mathbb P(\{\eta=\pm 1\})=1/2$. This has consequences for the predual $L^1(\Omega,\mathcal F,\mathbb P)/L^1(\Omega,\mathcal G,\mathbb P)$ of $E$ which the title of the article refers to.\N\NAs an application of this result, a generalization of Lyapunov's celebrated convexity theorem is obtained: If $\mathcal F$ is atomless conditionally to $\mathcal G$ and $f_1,\dots,f_n\in L^1(\Omega,\mathcal G,\mathbb P)$, then the set \N\[\N\{(\mathbb E(f_1\eta),\dots,\mathbb E(f_n\eta)): \eta\in \text{Ext}(\mathcal Z)\}\N\]\Nis a convex compact subset of $\mathbb R^n$.
This page was built for publication: A bipolar theorem only using independent random variables
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6913654)
