A simplified proof of a Bonferroni-type inequality by Tan and Xu and its extension (Q2639400)

Let \(A_ 1,A_ 2,...,A_ n\) be events on a given probability space, and let \(m_ n\) be the number of those \(A_ t\) which occur. Let \(S_ j\) be the j-th binomial moment of \(m_ n\). Guided by the known optimal lower bound on \(P(m_ n\geq 1)\) by means of \(S_ 1\) and \(S_ 2\), \textit{X. Tan} and \textit{Y. Xu} [Stat. Probab. Lett. 8, No.1, 17-20 (1989; Zbl 0672.60030)] developed a set of inequalities on \(P(m_ n\geq 1)\) by linear combinations of \(S_ 1,S_ 2,...,S_ r\), \(2\leq r\leq n\), which coincides with the optimal one for \(r=2\). The present author gives a new proof for an extended version of the inequality of Tan and Xu in which the coefficients are appropriately modified to obtain inequalities on \(P(m_ n\leq K)\) for arbitrary K. When \(K=0\) (write \(P(m_ n\geq 1)=1- P(m_ n=0)),\) these coefficients become those proposed by Tan and Xu. Reviewer's remarks: 1. In formula (1), the author misquotes the result of Tan and Xu, but in his Remark 1 on p. 256 the author states the correct form of the coefficients. 2. The upper inequalities on \(P(m_ n=0)\) by Tan and Xu have recently been transformed into lower inequalities by the reviewer and \textit{Y. Xu} [Math. Proc. Camb. Philos. Soc. 107, No.3, 601- 607 (1990; Zbl 0672.60031)]. By the new iteration method developed by the reviewer and Xu the inequalities of the present paper can also be transformed into lower inequalities.