You are on page 1of 5

A Characterization of the Power Function Distribution

Author(s): M. Ahsanullah and A. B. M. Lutful Kabir


Source: The Canadian Journal of Statistics / La Revue Canadienne de Statistique, Vol. 2, No. 1
(1974), pp. 95-98
Published by: Statistical Society of Canada
Stable URL: http://www.jstor.org/stable/3314967 .
Accessed: 15/06/2014 06:03
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Statistical Society of Canada is collaborating with JSTOR to digitize, preserve and extend access to The
Canadian Journal of Statistics / La Revue Canadienne de Statistique.

http://www.jstor.org

This content downloaded from 188.72.126.181 on Sun, 15 Jun 2014 06:03:17 AM


All use subject to JSTOR Terms and Conditions

La Revue Canadienne de Statistique


Section D: Notes et
rubrique des 6tudiants

The Canadian Journal of Statistics


Section D: Notes & Students' Corner
Vol. 2, No. 1, 1974, pp. 95-98

A CHARACTERIZATION
OF THE POWER
FUNCTIONDISTRIBUTION
by
M. AHSANULLAH
Protection
Branch, Ottawa

Health

and
A.B.M. Lutful KABIR
Carleton University
Key Words
and
Phrases:

Characterization, power function distribution, order


mean value theorem.
statistics,
independent statistics,

ABSTRACT
Suppose that in a random sample of size n from a
with probability
the order
function
population
density
f(x),
***
< X(n). It is proved that a
statistics
are X(1) < X(2) <
and sufficient
for f(x) to be the density
condition
necessary
function
of the power function
distribution
is that the
<
<
X(r/X(s)
statistics
and X
s
t< n) are independent.
(1 r
_
1.

INTRODUCTION

Let X(1) < X(2) < ... < X(n) denote the order statistics
in a random sample of size n from an absolutely
continuous
(with
to Lebesgue measure) distribution
function
F(x). Rogers
respect
a characterization
of the power function
dis[4] considered
tribution

(1)

)f(x)=
f~

c
-1
c-I
,
bc
= 0

O< x

b,

b > O, c > 0,

otherwise,

-95-

This content downloaded from 188.72.126.181 on Sun, 15 Jun 2014 06:03:17 AM


All use subject to JSTOR Terms and Conditions

ANDKABIR
AHSANULLAH
based on the independence
of the statistics
and
X(r/X(S)
X(s) (1 r < s < n) with s = r+l. Ahsanullah
extended
this
[2]
r
<
to arbitrary
result
and s (1 r s < n). In this paper we
_
further
extend the result
and give a necessary
and sufficient
based on the independence
condition
of the statistics
X(r)/X(s)
and X(t)(1 r < s < t n).
2. CHARACTERIZATION
OF THE POWERFUNCTIONDISTRIBUTION
Let X be a positive
THEOREM.
continuous
(with
absolutely

and bounded random variable


with
to
dismeasure)
respect
Lebesgue
<
and sufficient
tribution
function
b. A necessary
F(x), 0 x
_
the power function
distribution
as
condition
that X follows
<
<
s
r
is
for
s
t
t
the
that
some
and
n)
r,
given by (1)
(1 <
<
statistics
are independent.
X(r)/X(s) and Xt)

Proof.

is
The proof of the necessary
condition
To prove sufficiency
we need the following

trivial.
lemma which has

been proved in [2].


to
Let F(x) be an absolutely
continuous
(with respect
LEMMA.
=
and
let
distribution
function
measure)
Lebesgue
q(x,y)
F(xy)/F(y).
If for some r and s (1 r < s < n), the function
_
)

[q(x,y)] r-l[lq(x,y)]s-r-1
is independent
then q(x,y) is

of y for all y for which


of x alone.
a function

Proof of sufficiency:
we obtain

the

Letting
density

joint

h(u,v,w) = Kl[F(uvw)

U = X(r)/X(s), V = X(s)/X(t)
of U, V and W as
F(vw)-F

-li

[1-F(w)]n-tf(uvw)f(vw)
= 0 ,

F(y) > 0 and 0 < x < 1,

(uvw)]-r-[F(w)-F(w)t-s-1

f (w)vw2,

0<u<l,

otherwise,

where
K1

and W= X(t)

=
n!/[(r-1).(s-r-1)(Ct-s-1)!.(n-t)']
-96-

This content downloaded from 188.72.126.181 on Sun, 15 Jun 2014 06:03:17 AM


All use subject to JSTOR Terms and Conditions

0<v<1,

0<4wb,

POWERFUNCTIONDISTRIBUTION

Let kUW(u,w) be the joint density


of U and W and hw(w) the
of W. Since U and W are independent,
the marginal
density
of U is
density
h
(2)

hU(u) =

u(u,w)
hw()

KP

, 0 < u < 1 ,

T1(u, v,w) T2(u,w)dv

= 0 ,

otherwise,

where
= [q(u,vw) ]r-

T1(u,v,w)
=

T2(v,w)

= F(xy)/F(y)

q(x,y)

It

is

given

easy

to see

W = w where

h(u)

(t-1)!/[(r-1)!(s-r-1)!(t-s-1)!].

the conditional
Thus

p. 143] in

[3,

f10 T2(v,w)du

density

of

and

= KT1(u, ,w) , 0 < u < 1,

K = K2/K3. By the

independent

y = 8w,

q(u,Ow) = m(u),

of q(u,Gw),

a function

say,

we thus

F(uew) = F(ew)m(u)

of

independence

of W, and so by the

lemma with

only

of

u.

U and

W,

x = u and
From the

have

, 0 < ew < b,

0 < u < 1.

But this is a version


of Pexideris
[1, p. 1411. Since
equation
F(x) is a distribution
function
for all
continuous
absolutely
<
x
that
x, 0
< b, we may conclude
F(x) = axc , 0 < x
for

some

sufficiency

constants

is

b ,

c and a,

where

c > 0 and

a = 1/bC.

established.

-97-

This content downloaded from 188.72.126.181 on Sun, 15 Jun 2014 06:03:17 AM


All use subject to JSTOR Terms and Conditions

Thus

V
the

by applying

(2), we have

otherwise,

hU(u) is
definition

theorem

= K2T1(U, 6,w)

0 < e < 1

K3 = (t-1)./[(s-1)!(t-s-1)'].

= 0 ,
where

K2

K3T2(v,w) is

mean value

generalized

and

that

q(vw)

t-s-1

-1-q(v,w)

[q(v,w)]

au q (u,

-q(u, vw)

AHSANULLAH
ANDKABIR
ACKNOWLEDGEMENT

useful

The authors
comments.

are grateful

to the editor

and the referee

for

RESUME
Dans cette note, il est prouve qu'une condition
nec6ssaire
soit la fonction
de
et suffisante
de densit6
pour qu'une fonction
est
de
la
distribution
fonction
de
d'une
puissance
que
densit4
r < s < t <n soient
et X
les statistiques
inde(t),
X(r)/X(S)
de l'ordre
pendantes,
X(iW) tant par exemple la i-eme fonction
des observations.
REFERENCES
Lectures on Functional Equations and Their
New York.
Applications. Academic Press,

J. (1966).

[1]

Aczel,

[2]

Ahsanullah,

[3]

Olmsted,

[4]

Rogers,

of the power funcA characterization


M.(1973).
tion distribution.
Communicationsin Statistics,
2,
259-262.
J.M.H. (1959).
Real Variables.
New York.
Inc.,

Appleton-Century-Crofts

G.S. (1963). An alternative


proof of the characterization of the density
AxB . Amer. Math. Monthly, 70,
857-858.
Dr. M. AhsanuZZllah
Health Protection Branch
355 River Road, 4th floor
Ottawa, Ontario, CANADA
and
Professor A.B.M. Lutful Kabir
Department of Mathematics
Carleton University
Ottawa, Ontario, CANADA

Received 22 October 1973


Revised 1 April 1974

-98-

This content downloaded from 188.72.126.181 on Sun, 15 Jun 2014 06:03:17 AM


All use subject to JSTOR Terms and Conditions