Sharp upper bounds for the expected values of non-extreme order statistics from discrete distributions |
| |
Authors: | Fernando López-Blázquez Tomasz Rychlik |
| |
Institution: | 1. Facultad de Matematicas C/Tarfia, University of Sevilla, 41012 Sevilla, Spain;2. Institute of Mathematics, Polish Academy of Sciences, Chopina 12, 87-100 Toruń, Poland |
| |
Abstract: | We consider the problem of determining sharp upper bounds on the expected values of non-extreme order statistics based on i.i.d. random variables taking on N values at most. We show that the bound problem is equivalent to the problem of establishing the best approximation of the projection of the density function of the respective order statistic based on the standard uniform i.i.d. sample onto the family of non-decreasing functions by arbitrary N -valued functions in the norm of L2(0,1) space. We also present an algorithm converging to the local minima of the approximation problems. |
| |
Keywords: | primary 60E15 62G32 secondary 41A50 62D05 |
本文献已被 ScienceDirect 等数据库收录! |
|