sdtinalton 23.

34
Stas see we say that the statistic
this cas
converges stochastically to 0'as

Examplel
consishent
is
estfmator pyoblems.
mple mean is aa consistent estimator of the population mean
The
sample mean
Solution:

Let X1,2 , be a random sample from a population whose

unknownmnean is

F-AtX2t.
n

By Chebychev's inequality,

P-4s)sE(X-

(tie.) P - s e s

FimpX-)=0
n

estimator of .
.X is a consistent

Example 2 estimator ofthe population

consistent
variance is a
Show that the sample

variance 2.

Solution:
By Chebychev's inequality
E(s2-2 .(0)
P-o<)s*
E(2)2-202E(s2)+ o
E(2-2)2 =

distribution of s2
It is known from the sampling

ElG) E )
 23.35
Mathematical Stat

E(2-2)2

ptogg uv -1-2 +2n+r2

n2

lim
n>o
E-)-0
Therefore from (1), lim p(\s2 - o2|>e) = 0

i Hence 2 is a consistent estimator of o2,

Example 3
ns
Showthat is a consistent
estimator of g2.
Solution: lo 01mes atjas

jilatu
2no2
n-1

n - 1 ) 2Elo4)-2ng? E ) + o
n

n-1

On-1 o2no (-1)

1 h
AstimatioN 23.36
-2+1

AA

im Var -o1-1)=0 -(a . (1)

By chebychev's inequality,

ns
Var
n

lim P e using (1)

n n-1

is a consistent estimator of o2

Note:

ns also a consistent estimator of o2

is an unbiased estimator of a and
n-1

Theorem:
then 6, is a consistent estimator
If E(e )->0 and Var(0,)>0 asn o ,
of .

Proof: By Chebychev'sinequality, nb

Pp.-zie.)»tFaOJ|
1 or any n,
 23.37
choose =kVar
MathematicalsStatis
ThenP , -E(©,)}<e]>l- . (1)

le--(e,-E(0.))+E(0,)-
so,-B(.)+|(0,)-4
Since E()->0as n o given e2>0 there exists positivenumber Ne
Nsucht

EO,)-<e ifn> N -b-1 -

: 10,-6<0, -E(O,)}+ E if n>N everiavdede
va
From(1) P(2, -0<e +s)>1Pare).
Further since Var(0,)>0 when n-00
Var(0,) <E; when n> N

PO,-0<(G+5)>1-
choosing e +=6=n

to otesnoi we get Pn6|

KE)17 sasidnu n ei
limP,-0=
n-o A 2

lim pp,-e=
n0

G, 1S a Consistent estmator ofAR

2olagdori 00
 s2=
the
sample
variance (X-X is aconsistent estimator
Showthat
th

populationv a variance 2 where (X1,X2, X)is arandom sample from

lation
m a l populaton N ( o , 2 )
).
alntion
,2 9
We knowthat, E(r2)=
bis
Var()4
n-1

E(s)o and Var()>0forn.

consistent estimator of 2.
2 is a
Hence by the (

Example5 of the population

estimator
median is a consistent
Show that the sample
normal population.
mean of a
Solution:
normal population
random sample of size n from
a
is a
If (X,X2,-X) median is formed tothe
function of the sample
for large n, the density

proportionalto-2n.-0
ordinate equal to v2no
median
where S is the population

Var(X)= T
E(X)=0 and 2nT

Var Oasn-0

populationmean.
.X is consistent estimator ofthe
 23:39 Mathematie
Example 6
If X,2 dn are random observations on a Bernoull
ityvariable
1-P. Show
y

and the value 0 with probability 1-p, sh

value 1 with probability p
p(l-p)
is a consistent estimator of

Solution:
AA2,X, are independent and
identically istributet Bemóu
distributet a

variables with parameter n and p.

..+X, is a Bermoulli variable with pDar
Then T-X+2* amete
n and p.
E(7)=np and Var(T)=rpg

F.
ET)E
ECX)np=p

Var(X)=Va toHoiotis
Var(T)

Since E(X)>p and VarX0 as->o, isa consistent estimatorof p.

Also EX -Fo-
n
X(1-X) being a polynomial in is a continuous function of
 23 40
s a consistent estimator of p, byInvariance property of consistent
Since
S i n o

r1-X isalsoa consistent estimator of p(1-p)

estimator,

Srample
-00X0
he
chowthat from the Cauchy population
-
sanmplemeani a ienot aconsistant estimatorbut sample median is aconsistentestimator
Solution:

n bea random sample from the Cauchy population. Its

Let XX2
probabilitydistribution is

S,p)=-

put X-=z,

AlHe- E

2tane 23

cannot tend to zero

as n , Hence
that Pl>e
It is independent ofn
so

not a consistent
estimator of
the sample mean is

is 4 for the given caucny

distribution of median
The sampling
population.
23.41 Mathematical Statist
Here E(X)=4 are VarX 4n1

n-o
lim Var(X)=lim =0
no 4n

xis a consistent estimator of .

 and
where & pe
n1A2.n) does not involve 0.

Exampl boilev bls

that is a sufficient estimator of
Chow the mean H ofa normal population
nown
with the known variance a stotueSntee eo yst o
Solution:

The likelihood function of the sample is given by olqm?

e 20e 9
(ov2T

Now -
i=l
w - - -
isl 1012 2 i w
o sre v 2nomanie boieuro
2
b +2)4+2-A

- +0+x

E-+{-u

1
L
o2

oV2n