On statistical information of extreme order statistics, local extreme value alternatives, and Poisson point processes (Q1323139)

The paper discusses the role of extreme order statistics in the general setup of statistical experiments in \textit{L. LeCam's} sense [Asymptotic methods in statistical decision theory. (1986; Zbl 0605.62002)]. Under the assumption of monotone likelihood ratios, it is shown that a sequence of experiments is asymptotically Gaussian as sample size increases if the \(k\) extreme order statistics asymptotically do not contain any information for any fixed \(k\). When the data is made up of the only the \(k\) extreme order statistics, representation for the limit experiment by Poisson point processes is given. A new extreme value model given by local alternatives is proposed. It is used to find asymptotically optimal tests within certain threshold models.