No label defined (Q4239170)

The authors prove the Kantorovich inequality for conditional expectations inspired by the recent work due to \textit{V.~Pták} [Am. Math. Mon. 102, No. 9, 820-821 (1995; Zbl 0856.26011)]. They also give a necessary and sufficient condition that the equality holds in the Kantorovich inequality of the following integral form: Let \((\Omega,{\mathcal F},p)\) be a probability measure space. For any complex valued random variable \(X\) with \(0<m\leq |X|\leq M\)\ a.s., we have the following inequality \[ \left|\int Xdp\int \frac{1}{ X }dp\right|\leq\frac{ (m+M)^{2} }{4mM}. \] The equality is attained if and only if there exists \(0\leq\theta<2\pi\) such that \(p(\{X=me^{i\theta}\})=p(\{X=Me^{i\theta}\})=1/2\) (if \(m<M\)) or 1 (if \(m=M\)). The result is applied to an inequality of normal operators on Hilbert spaces.