The Fisher information of curved exponential families and the elegant Kagan inequality (Q2117865)

Today, information processing permeates all areas of our daily life. The overview of its history can thus only be presented in a spotlight manner and should above all encourage readers to seek out more detailed literature. Life without information processing is not possible. So energy, matter, and information are the most important basic concepts in the natural sciences. Indeed, \textit{A. M. Kagan} [Teor. Veroyatn. Primen. 30, No. 4, 783--786 (1985; Zbl 0582.60022)] discovered in 1985 a remarkable inequality on Fisher information. Curved exponential families [\textit{E. Angetun}, ``On curved exponential families'', Preprint] are so general objects that they seem to have no interesting universal properties. After recalling Fisher's models and Fischer's information, exponential families (natural exponential families, Fisher information of a natural exponential family, general exponential families, curved exponential families) and Kagan inequality, the author eliminates the artificial role of general exponential families by going to the core of the proof of the results and by dealing only with a natural exponential family and the curved one issued from it, and he illustrates by examples the original Kagan inequality [Kagan, loc. cit.]. The author gives a modern presentation of this result and examples, comparing in particular noncentral and central Wishart distributions [\textit{G. Letac} and \textit{H. Massam}, J. Multivariate Anal. 99, No. 7, 1393--1417 (2008; Zbl 1140.62043); J. Multivariate Anal. 163, 96--110 (2018; Zbl 1386.60068)] at the end of the paper. For the entire collection see [Zbl 1482.94007].