Characterization based on conditional expectations of adjacent order statistics: A unified approach

From MaRDI portal





Let \(X_{r:n}\) denote the \(r\) th order statistic of a random sample of size \(n\) from an arbitrary distribution function \(F\). Assume \(h\) to be a strictly monotonic continuous function and define the conditional mean functions (cmf) \[ A_F (x)=E[h(X_{r:n}) | X_{r+1:n} = x]\quad \text{and}\quad B_F (x)=E[h(X_{r+1:n}) | X_{r:n} = x]. \] In this paper, \((i)\) necessary and sufficient conditions for an arbitrary function to be a cmf are given, \((ii)\) an inversion formula that recovers \(F\) from either \(A_F\) or \(B_F\) is presented and \((iii)\) equivalence of \(F_n @>w>> F\) and \(A_{F_n} @>w>> A_F\) is established. This work generalizes and unifies past work on cmf's for discrete and continuous distributions. These results are applicable, for example, to the characterization of mixtures of discrete and continuous distributions.











This page was built for publication: Characterization based on conditional expectations of adjacent order statistics: A unified approach

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4226298)