A particular case of correlation inequality for the Gaussian measure (Q1577745): Difference between revisions

The paper deals with the following conjecture, which is due to \textit{S. Das Gupta}, \textit{I. Olkin}, \textit{L. J. Savage}, \textit{M. L. Eaton}, \textit{M. Perlman} and \textit{M. Sobel} [Proc. 6th Berkely Symp. Math. Statist. Probab., Univ. Calif. 1970, 2, 241-265 (1972; Zbl 0253.60021)]. Let \(A\) and \(B\) be two symmetric convex sets; if \(\mu\) is a centered Gaussian measure on \(\mathbb{R}^n\), could we say that \(\mu(A\cap B)\geq \mu(A)\mu(B)\)? In the paper, the conjecture is proved if \(A\) is a centered ellipsoid and \(B\) is simply symmetric. Two proofs are given, the first one is based on the log-concavity result of \textit{A. Prékopa} [Acta Sci. Math. 34, 335-343 (1973; Zbl 0264.90038)], and the second one relies on the comparison of semigroups. As a corollary the authors obtain a comparison of the moments of a measure with an even, log-concave density with the moments of \(\mu\).