Improved Hölder and reverse Hölder inequalities for Gaussian random vectors (Q2346052)

Consider a jointly Gaussian vector which is composed of \(m\) subvectors \(X_i\) of possibly different size with overall covariance matrix \(T\). Let \(P\) be a block diagonal matrix with blocks that describe the internal covariances of the subvectors multiplied with constants \(p_i\). The authors prove: {\parindent=6mm \begin{itemize} \item[(i)] If \(P - T\) is positive semi-definite then a generalized Hölder-inequality holds. \item [(ii)] If \(T - P\) is positive semi-definite then a generalized reverse Hölder-inequality holds. \end{itemize}} It is shown that the classical Hölder inequality and the reverse Hölder inequality are special cases. The authors discuss in detail the connection of their results to different inequalities and related topics from the literature, among others: Prekopa-Leidler inequality, sharp Young inequality; reverse Young inequality, Braskamp-Lieb inequality, Barthe inequality, Gaussian hypercontractivity, and an entropy inequality.