This article was downloaded by: [Purdue University]

On: 17 January 2015, At: 00:46

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41
Mortimer Street, London W1T 3JH, UK

Journal of the American Statistical Association

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uasa20

On the Bounds of the Range of Order Statistics

a
Douglas M. Hawkins
a
Department of Statistics , University of the Witwatersrand , Johannesburg , South Africa
Published online: 05 Apr 2012.

To cite this article: Douglas M. Hawkins (1971) On the Bounds of the Range of Order Statistics, Journal of the American Statistical
Association, 66:335, 644-645

To link to this article: http://dx.doi.org/10.1080/01621459.1971.10482325

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the
publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or
warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed
by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with
primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings,
demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly
in connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is
expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
 @ Journal of the American Statistical Asrodation
September 1971, Volume 66, Number 335
Theory and Methods Section

On the Bounds of the Range of Order Statistics

DOUGLAS M. HAWKINS”

~~

Let XIXP .. * x,, be any set of numbers satisfying x i =0 x? = 7. There ore Define
bounds on the ronge o f x i when the set { x i ) is ranked ..
( 0 ) by ~ 1 2 x 2 2 * 2x,

and ( b) by I xi1 2 I x2l 2 - * * 2 I x.1. There bounds ore given. The applicobilify r = (y - Z)/s;
to finite populotionr is pointed out.
then
1. INTRODUCTION AND RESULTS r 2 b ( j , 4. (1.5)
Let b(1, n), b(2, n) - b(n, n) be a set of n numbers Assume that the xi are chosen in such a way that b ( j , n)
satisfying attains its minimum m(j,n). Then (1.5) shows that

C L l b ( i , n) = 0 CZ,byi, n) = 1 (1.1)
b(1, n) 2 b(2, n) 2 2 b(n,n).
Downloaded by [Purdue University] at 00:46 17 January 2015

* * f
(1.2)
Let T = m ( j , n) if y = 2,.

M ( i , n) = max b(i, n) Clearly no other choice of the set (xi] or the value y
m ( i , n) = min b ( i , n), decreases this value of r . Thus min(r) = m ( j , n) the mini-
mum being over all { z i ) and y satisfying the ranking
the extrema being taken over all sets of b(i, n) satisfying constraints. From the augmented set, we can compute a
(1.1) and (1.2). set { b ( i , n + l ) 1,
in which b ( j , n + l ) corresponds to y.
Theorem 1 : It is known [ 2 ] that b(j, n + l ) is given by
1
(a) m,(l,n) = -
d n ( n - 1)
a strictly increasing function of r . Thus b ( j , n + l ) is
minimized when r is minimized. So

To get m(n4-1, n+l), let y<x, tend to - a. Then r

- - - 00 and b ( n + l , n + l ) tends to its lower bound. Thus

nr
Proof: Let x 1 2 x 2 2 - 22, be a set of real numbers, m(n + 1,n + 1) = I--
lim
- (n + 1 ) d l + nr”(n + 1)
not all identical.
Part (a) of the theorem follows by induction, and the fact
that for n = 2 -b(1, 2) =b(2, 2) =1/d/2. Part (b) follows
from the fact that
M ( i , n) = - m(n - i + 1,n),
b(i,n) = (xi - 4 / s . as can be seen by multiplying all xi by -1.
The b ( i , n) satisfy (1.1) and (1.2) above. Theorem 2: Let the set { b(i, n) ] satisfy conditions
Let the set {xi) be augmented with some number y, and (l.l), and also
let y have r a n k j s n , in the augmented set: i.e.,
Y 2 xi (1.4)
I b(l,n) I 2 I b(2,n) I 2 . . 2 I b(n, n) [ (1.7)
*

* Douglaa M. Hawkins is senior lecturer. Department of Statistics, University

I
Let M’(i, n) = max b (n, n) I over all sets satisfying (1.1)
of the Witwatersrand, Johannesburg, South Africa. and (1.7). Then
644
 Bounds on Order Statistics 645

M‘(i, n> = &- i even

b ( i , n) =
-f
__
Xi

dn.
.

Theorems 1 and 2 then show the constraints on the b(i, n)

The proof of this theorem is given in an appendix by
Scott to [l].The proof is constructive, showing for what
1
value of the set { b(i, n) this maximum is attained.
2. APPLICATION
Let xl> x2 > . . 2 x, be a finite population with mean
$, and non-zero variance uz.
Let
Downloaded by [Purdue University] at 00:46 17 January 2015
imply bounds on the deviations of the xi from 5.
Samuelson [3] has shown that Theorem 2 yields an in-
equality sharper than that of Chebyshev. It is of interest
to note that there is a discrepancy between Samuelson’s
second theorem and the corresponding Theorem 2 of this
article.
REFERENCES
[l] Pearson, E. S. and Chandra, Sekar C., “The Efficiency of
Statistical Tools and a Criterion for the Rejection of Out-
lying Observations, ’’ Biometrika, 28 (1936), 308-20.
[2] Quesenberry, C. P. and David, H. A., “Some Tests for
Outliers,” Biometrika, 48 (1961), 379-90.
[3] Samuelson, P. A,, “How Deviant Can You Be?” Journal of
the American Statistical Association, 63 (December 1968),
1522-5.