Conditional quantiles of Gaussian measures in Hilbert space (Q1269992)

Let \(H\) be a separable real Hilbert space with inner product \(\langle\cdot,\cdot\rangle\). Let \(\mu\) be a Gaussian measure on \(H\) with mean vector \(m\) and correlation operator \(B\). Let \(H_{+}\) and \(H_{-}\) be Hilbert spaces obtained by using a construction from the theory of equipped Hilbert space: \(H_{+}\subset H\subset H_{-}\), \(B^{1/2}H_{-}=H\), \(B^{1/2}H=H_{+}\), \(\langle\cdot,\cdot\rangle_{-}=\langle B\cdot,\cdot\rangle\), \(\langle\cdot,\cdot\rangle_{+}=\langle B^{-1}\cdot,\cdot\rangle\) [see \textit{A. V. Skorohod}, ``Integration in Hilbert space'' (1974; Zbl 0307.28010)]. An orthonormal basis \(\{f_k\}\) in \(H\) is said to be minimal with respect to the norm \(| \cdot| _{-}=\sqrt{\langle\cdot,\cdot\rangle_{-}}\), if for any \(k=1,2,\ldots,f_k\) does not belong to the closed linear hull of the vectors \(f_n\), \(n=1,2,\ldots\), \(n\neq k\) with respect to \(| \cdot| _{-}\). After studying minimality of an orthonormal basis \(\{f_k\}\) in \(H\) with respect to \(| \cdot| _{-}\), the author considers the quantile of order \(p (0\leq p\leq 1)\) of the conditional distribution function of the Gaussian random functional \(\langle f_k,\cdot\rangle\) given Gaussian random functionals \(\langle f_1,\cdot\rangle, \ldots,\langle f_{k-1},\cdot\rangle, \langle f_{k+1},\cdot\rangle,\ldots\) on \((H,\mu)\). The author proves that if \(\{f_k\}\) is minimal with respect to \(| \cdot| _{-}\) and \(m\in H_{+}\), then the quantiles above are uniquely determined by their explicitly calculated restrictions onto the kernel \(H_{+}\) of the Gaussian measure \(\mu\).