On bounds for the range of ordered variates II
Letx1, ⋯,x n be real numbers with∑ 1 n x j =0, |x 1 |≦|x 2 |≦⋯≦|x n |, and ∑ 1 n f(|x i |)=A>0, wheref is a continuous, strictly increasing function on [0, ∞) withf(0)=0. Using a generalized Chebycheff inequality (or directly) it is easy to see that an upper bound for |x m | isf −1 (A/(n−m+1)). If (n−m+1) is even, this bound is best possible, but not otherwise. Best upper bounds are obtained in case (n−m+1) is odd provided either (i)f is strictly convex on [0, ∞), or (ii)f is strictly concave on [0, ∞). Explicit best bounds are given as examples of (i) and (ii), namely the casesf(x)=x p forp>1 and 0<p<1 respectively.
AMS (1970) subject classificationPrimary 62G30, 26A51 Secondary 26A87
Unable to display preview. Download preview PDF.
- Beesack, P. R.,On bounds for the range of ordered variates. Univ. Beograd. Publ. Elektrotehn. Fak. Ser. Mat. Fiz.412–460 (1973), 93–96.Google Scholar
- Boyd, A. V.,Bounds for order statistics. Univ. Beograd. Publ. Elektrotehn. Fak. Ser. Mat. Fiz.357–380 (1971), 31–32.Google Scholar
- Hawkins, D. M.,On the bounds of the range of order statistics. J. Amer. Statist. Assoc.66 (1971), 644–645.Google Scholar
- Loève, M.,Probability theory,2nd ed. Van Nostrand, New York, 1960.Google Scholar
- Pearson, E. S. andChandra Sekar, C.,The efficiency of statistical tools and a criterion for the rejection of outlying observations. Biometrika28 (1936), 308–320.Google Scholar
- 6]Samuelson, P. A.,How deviant can you be? J. Amer. Statist. Assoc.63 (1968), 1522–1525.Google Scholar