No label defined (Q3843201)

This article, Chapter 10, has deep interrelations with Chapter 7 of this volume written by \textit{N. Balakrishnan} and \textit{K.S. Sultan} [see the review Zbl 0907.62054 in this Zbl volume]. It demonstrates how general and specific recurrence relations and identities for moments of order statistics and their modifications can be used to characterize the main types of probabilistic distributions. In Section 2 the author mentions but does not review characterizations by moments and moment differences. Results of this type are mainly based on complete sequences of functions. Several of these sequences are cited. Section 3 contains the characterization of exponential distributions due to Z. Govindarajulu. Section 4 deals with the class \({\mathcal F}\) of distribution functions \(F\), where \(F\in{\mathcal F}\) is determined by the derivative of its pseudo inverse function according to the equation \((F^{-1})'= d^{-1}t^p(1-t)^{q-p-1}\) with a constant \(d>0\) and integers \(p,q\). This general approach due to the author admits to summarize and generalize a number of isolated results and provides an insight into the structural properties and relationships of several probabilistic distributions such as exponential, Burr, Weibull, Pareto, logistic and related distributions. Several characterization results which are based on a single identity in contrast to the one deduced by complete function sequences are cited in Section 5. Finally, Section 6 presents a review of some results based on product moments of order statistics including Govindarajulu's results for normal distributions. Other approaches to characterizations by means of order statistics are reviewed in Chapters 8 and 9 of this volume. For the entire collection see [Zbl 0894.00024].