Stochastic orders of the sums of two exponential random variables (Q5933625)

The aim of the paper is to investigate the sums of two exponential random variables from the point of view of likelihood ratio (LR) order. Let \(X\) be the sum of two independent exponential random variables with rates \(g_1\) and \(g_2\), and let \(Y\) be the sum of two independent exponential random variables rates \(h_1\) and \(h_2\). Let also (without loss of generality) \(g_1\geq g_2\) and \(h_1\geq h_2\). The main result of the paper provides a complete comparison of \(X\) and \(Y\) in the sense LR order, under which conditions the LR order does (not) exists, and determines LR order if it exists. Namely, the author proves the following Theorem: Given that \(g_1+ g_2\geq h_1+ h_2\), if \(g_1+ g_2> h_1+ h_2\) and \(g_1> h_1\geq h_2> g_2\), then \(X\) and \(Y\) are not comparable in the LR order; if \(g_1+ g_2= h_1+ h_2\) and \(g_1> h_1\geq h_2> g_2\), then \(X\geq_{\text{LR}}Y\); otherwise \(X\leq_{\text{LR}}Y\). Various applications of stochastic ordering of random variables may be found in such fields as economics, system reliability, machine ordering, etc.