Ordering results between two multiple-outlier finite \(\delta\)-mixtures (Q6589435)

For \(i=1,\ldots,n\), let \(\overline{G}(\cdot;\gamma_i)\) denote a survival function depending on a parameter \(\gamma_i\). Given positive mixing proportions \(r_1,\ldots,r_n\) with \(r_1+\cdots+r_n=1\), and \(\delta\in\mathbb{R}\), the survival function of the corresponding \(\delta\)-mixture model is defined by\N\[\N\overline{G}_\delta(t)=\begin{cases} \left(\sum_{i=1}^nr_i\overline{G}^\delta(t;\gamma_i)\right)^{1/\delta} & \text{if }\delta\not=0\,,\\\N\prod_{i=1}^n\overline{G}^{r_i}(t;\gamma_i) & \text{if }\delta=0\,. \end{cases}\N\]\NThe authors consider the special case in which there is \(n_1<n\) such that \(\gamma_i=\gamma\) for \(i\leq n_1\) and \(\gamma_i=\gamma^\prime\) for \(i>n_1\). They establish various stochastic ordering results for the comparison of two such models, using the usual stochastic order and the reversed hazard rate order. Several examples are used to illustrate these results.