Random sets and random fuzzy sets as ill-perceived random variables. An introduction for Ph.D.\ students and practitioners (Q2253780)

This is a very interesting and carefully written research exposition. Already the informative introduction clearly presents the aim of the book. At first, the authors discuss two approaches for random sets. In the (mainstream) approach by \textit{D. G. Kendall} [in: Stochastic Geom., Tribute Memory Rollo Davidson, 322--376 (1974; Zbl 0275.60068)] and \textit{G. Matheron} [Random sets and integral geometry. New York, etc.: John Wiley \& Sons (1975; Zbl 0321.60009)], a realisation of a random set is a (precisely located) set in an area of interest (so-called ontic approach). But there is another approach: a random set can be a natural model for an ill-known random variable. A random set is then a natural representation of a set of possible variables (so-called epistemic approach). This view leads to Dempster's upper and lower probabilities and to Shafer's evidence theory. This distinction between ontic and epistemic approach is highlighted also for random fuzzy sets (rfs). The ontic approach goes back to \textit{M. L. Puri} and \textit{D. A. Ralescu} [J. Math. Anal. Appl. 114, 409--422 (1986; Zbl 0592.60004)]. Here, the realisation of a rfs is indeed a fuzzy set. The epistemic approach [\textit{H. Kwakernaak}, Inf. Sci. 15, 1--29 (1978; Zbl 0438.60004); \textit{R. Kruse} and \textit{K. D. Meyer}, Statistics with vague data. Dordrecht, etc.: D. Reidel Publishing Company (1987; Zbl 0663.62010)] however models a rfs as the imprecise perception of an ill-known classical random variable. The variance of a rfs can be seen as an indicator of the different approaches: the variance of a rfs is a precise number in the ontic approach, but a fuzzy set in the epistemic one. In this book, the authors present the basics of the epistemic approach for both, random sets (Chapter 2) and random fuzzy sets (Chapter 3). Using some tools of imprecise probability theory they bridge the gap between the works of Kwakernaak-Kruse-Meyer tradition and \textit{A. P. Dempster}'s work in upper and lower probabilities [Ann. Math. Stat. 38, 325--339 (1967; Zbl 0168.17501)]. Typical catch words are, e.g., the acceptability function of a rfs and the fuzzy probability envelope. The latter one represents the available (imprecise) information about the (unknown) original probability measure. The text is written not in the definition-theorem-proof manner, but the authors use a lot of carefully chosen examples to discuss their ideas. Somewhat misleading seems to be the subtitle of the book: An introduction for PhD students and practitioners. For the reviewer, this book is a methodological research discussion.