Sufficient conditions for some transform orders based on the quantile density ratio (Q2241496)

Let \(X\) and \(Y\) be two non-negative random variables with distribution functions \(F\) and \(G\) and density functions \(f\) and \(g\), respectively. It is said that \(X\) is smaller than \(Y\) in the \textit{convex transform order} if the \textit{quantile densities ratio} \(\displaystyle \frac{f\left(F^{-1}(p)\right)}{g\left(G^{-1}(p)\right)}\) is increasing in \(p\in(0,1)\), where \(F^{-1}\) and \(G^{-1}\) are the generalized inverses of \(F\) and \(G\), respectively. The paper studies the situations where the convex transform order does not hold but some other weaker transform orders are satisfied. Under the assumption that there exists a point \(p^*\in(0,1)\) such that the quantile densities ratio is increasing over \((0,p^*)\) and decreasing over \((p^*,1)\) (therefore the quantile densities ratio is unimodal), necessary and sufficient conditions for the \textit{star-shaped} ordering and for the \textit{expected proportional shortfall ordering} are proved. Also, assuming that the quantile densities ratio has a finite number of modes, the authors provide necessary and sufficient conditions for the \textit{qmit ordering} and for the \textit{dmrl ordering}. The results about the comparison of the non-negative random variables in various transform orders are then summarized and commented. In addition, applications in reliability are provided. The theoretical results are illustrated by numerical examples.