Limiting multivariate distributions of intermediate order statistics (Q1062376): Difference between revisions

Let Z(m,n) represent the mth largest (reverse) order statistic in a random sample of size n from a distribution F. Further let \(m=m(n)\) be an intermediate sequence of (n) in the sense that m/n\(\to 0\) and \(m\to \infty\) as \(n\to \infty\). The author finds the asymptotic joint distribution of order statistics of the form Z([mt],n), where \(0<t_ 1<t_ 2<...<t_ k\) for various types of F. This is done by determining the weak limit of a process of the form \((Z([m,t],n)-\beta (mt,n))/\alpha (mt,n))\) for appropriate normalizing functions \(\alpha (.)>0\) and \(\beta (.)>0.\) The author indicates that the joint distribution of these order statistics may be used for making statistical inferences about the upper tail of F, but this question is not pursued any further.