Computing the moments of order statistics from nonidentically distributed phase-type random variables (Q631888)

The phase-type (PH) distribution is a distribution of the time until absorbtion for a finite-state Markov process. Let \(X_{1:n}\leq\dots\leq X_{n:n}\) be the order statistics of independent but not identically distributed PH random variables. The authors derive explicit formulae for the moments of \(X_{1:n}\) and \(X_{n:n}\) and a recurrence relation for the moments of \(X_{r:n}\). These results are based on the permanents representation for the distribution of \(X_{r:n}\) and the Laplace-Stieltjes transformation of the distribution. An application to stochastic activity networks is discussed.