Characterizations of stochastic ordering for non-negative random variables (Q6084754)

The authors establish the following characterization of the usual stochastic order between random variables \(X\) and \(Y\) supported on \([0,\infty)\) with density functions \(f\) and \(g\), respectively: Suppose \(l(x)=\log(f(x)/g(x))\) is a continuous function with at most two roots, such that \(l(x)<0\) for all \(x\) greater than the largest such root. Then \(X\) is stochastically smaller than \(Y\) if and only if \(l(0)\geq0\). Analogous characterizations are also established for the likelihood ratio order, the hazard rate order, and the reversed hazard rate order. The conditions in these characterizations are weaker than the concavity of \(l(x)\) required in the previous work of \textit{Y. Yu} [J. Appl. Probab. 46, 244--254 (2009; Zbl 1161.60308)].