Moment inequalities (Q1119252)

The author collects a number of results from the literature which show that the coefficients in the continued fraction expansion of the Stieltjes transform of a measure m(.) determine the moments \(u_ k\) of m(.) by an explicit rule. He utilizes this result in developing a method for improving on moment inequalities. The method is demonstrated for the special case of improving some Bonferroni-type inequalities. In this latter case, the basic idea is the following: Let \(A_ 1,A_ 2,...,A_ n\) be events, and let \(m_ n\) be the number of those A's which occur. Let \(u_ k=E(m^ k_ n)\). Assume that, with some coefficients \(c_ k\) and \(d_ k\), the Bonferroni-type inequalities \[ \sum^{s}_{k=0}c_ ku_ k\leq P(m_ n\geq r)\leq \sum^{s}_{k=0}d_ ku_ k \] have been established, and that \(u_{s+l}\) is either determined, or can be estimated, by \(u_ 1,u_ 2,...,u_ s\). Then, given \(u_ 1,u_ 2,...,u_ s\) only, one can write up the quoted Bonferroni-type inequalities up to \(s+l\), use the terms up to s, and estimate the contributions of \(c_{s+l}u_{s+l}\) and \(d_{s+l}u_{s+l}\) by means of \(u_ 1,u_ 2,...,u_ s\). The method is worked out in detail for some particular inequalities.