Inequalities for expected extreme order statistics (Q2483890)

This paper considers \(n\)-dimensional random vectors having independent integrable components and investigates suitable conditions to obtain bounds for differences of expected extreme order statistics. Section 2 contains the main results, providing one-sided estimates for differences of expected extreme order statistics in terms of marginal distributions, provided that such distributions fulfil the so-called Pivot conditions. The general results in Section 2 are then applied to special cases in the following sections. Section 3 compares the expected extreme order statistics of random samples from binomial and Poisson distributions having the same mean, while Section 4 is doing the same thing for random samples from Poisson and negative binomial distributions with the same mean. The final Section 5 applies the basic results from Section 2 to derive a method for obtaining lower (upper) bounds for expected maxima (minima) of integrable independent random variables in terms of the parent distribution function.