Preservation of some likelihood ratio stochastic orders by order statistics (Q5933607)

The aim of the paper is to extend results on order statistics related to the notions of likelihood ratio order and up (and down) shifted likelihood ratio order, introduced and studied by \textit{J. G. Shanthikumar} and \textit{D. D. Yao} [Stochastic Processes Appl. 23, 259-267 (1986; Zbl 0611.60047)] and \textit{M. Brown} and \textit{J. G. Shanthikumar} [Probab. Eng. Inf. Sci. 12, No. 4, 425-444 (1998; Zbl 0964.60014)]. Sections 1 and 2 of the paper introduce the necessary definitions of the likelihood ratio order, stochastic order, hazard rate order, dispersive order, up (and down) shifted likelihood ratio order for the order statistics from a sample of independent, identically distributed random variables. There are established and generalized results on the above-mentioned order statistics for (nonnegative) absolutely continuous random variables, having logconcave density functions. Section 3 of the paper compares order statistics, from two different samples, in the likelihood ratio and in the shifted likelihood ratio orders. Shifted orders are proved to be a very useful tool for establishing stochastic inequalities. For instance, the easy identification of the shifted likelihood ratio orders, combined with the fact that order statistics are liftimes of \(k\)-out-of-\(n\) systems, yield interesting bounds in the reliability theory.