M.

Riaz Lecturer Statistics The Islamia University of Bahawalpur RYK Campus- 1

Consistency
If we increase the sample size then our estimator is closer and closer to the parameter it is called
consistency. If estimate become closer to parameter then the variance of this estimate is
minimize. As n   then variance of the parameter is zero.
Or
An estimator is said to be consistent only if and only if with the increase in sample size or it
becomes closer and closer to parameter which is known as consistent estimator.
Or

An estimator ' ' , is said to be consistent if it tends to parameter ‘  ’.

As n  

Lim P[  _  ]  0
n 


Lim P[  _  ]  1 Condition-I
n 


Lim Var ( )  0 Condition-II
n

PROCEDURE

First we consider Lim P[  _  ]  1 and convert it as P[ z  ]  1. If it is true then consider
n


Lim Var ( ). Then it must be equal to zero.
n 

 Limn  Var ( )  0

If these conditions are satisfied then estimator are said to be consistent.

Q.1
Show that sample mean is consistent estimator of the parameter  .
If x ~ N ( ,1)

Solution
As X ~ N ( ,1)

X ~ N ( , 1
n)

i.e

E (X ) = E ( )  

Var( X )= Var ( )  1 n

Now we consider

Lim P[  _  ]
n

Replacing these values:

 M.Riaz Lecturer Statistics The Islamia University of Bahawalpur RYK Campus- 2


== Lim P[  _  ]
n

 1
Dividing by 1
n  x
n

 

= lim p[  ]
n 1 1
n n

  
 
= lim p[  n ] z 
n 1 1
n n


= lim p z  n 
n


= p z  
= p   z  

=1
So C-1 is satisfied
 1
Now lim v( ) lim  lim 0
n n n  n
As condition 1 and condition 2 are satisfied so sample mean is consistent estimator for ”  ”
Q.2
Show that sample mean is consistent estimator of the parameter 


If x~N  ,  2 
Solution

As x~N  ,  2  
 2 
x ~N   , 
 n 

i.e

E( x )= E(  )= 


2
Var( x )=v( ̂ )=  x =
n
We consider

Lim P[  _  ]
n

Replacing the values

= Lim P[ x   ]
n 


Dividing by
n
 M.Riaz Lecturer Statistics The Islamia University of Bahawalpur RYK Campus- 3

 
 x     
= Lim P   
n 
   
 n n 

 
 x    n 
= Lim P   
n 
   
 n 

 n 
= Lim P  Z  
n 
  

= Lim P Z  
n 

= Lim P   Z  

n 

=1

So C-1 is satisfied

Now
 2
Now lim v( ) lim  lim 0
n n  n  n

2
=

=0
As c-1 and C-2 are satisfied so sample mean is consistent estimator for  .

Q. 3
Show that from the random sampling of Cauchy distribution
1
f(x)=   x  
 (1  ( x   ) 2 )

Then sample mean is not consistent estimator for the parameter and sample median is a
consistent estimator for the parameter.

Solution

(For population mean)

1
As f(x) =   x 
 (1  ( x   ) 2 )
Put z= x - 
dz= dx
1
f(z)=   x 
(1  z 2 )

We consider

  
lim p     
n
 
 M.Riaz Lecturer Statistics The Islamia University of Bahawalpur RYK Campus- 4

= lim p x   
n
 
= lim p z 
n
 
= lim p  z 
n


f ( z )dz
=


 1
1 dz
=
  1 z2



1
= tan 1 ( z ) 
 

=
1

tan 1
()  tan 1 ( ) =  1

tan 1
()  tan 1 () 
2 tan 1 ()
1

As C-1 is not satisfied hence sample mean of Cauchy distribution is not consistant estimator for
.
Now we consider the median of Cauchy distribution
~
E( x )= 
~
2
Var( x )=
4n
Now we consider

~ 
   
lim p      = lim p x   
n
  n  
 

 ~ 
 x   
= lim p   
n
  
2 n 2 n
 

 2 n 
= lim p  z  
n
  

= lim p z  
n
 
= lim p   z  
n

=1
So condition 1 is satisfied
Now
~
2
lim v( x)  lim ( )
n  n 4n
 M.Riaz Lecturer Statistics The Islamia University of Bahawalpur RYK Campus- 5

2
=
4(  )

2
=

=0
As C-1 and C-2 are satisfied so sample median of Cauchy distribution is consistant
estimator for parameter  .
Q. 4

If X1,X2,X3,…,Xn be a random sample of size n from a normal distribution (  ,  2 )

Then show that:

1) x is a consistent estimator of 

2) s 2

 ( x  x) 2

consistant estimator of  2
n 1
Solution
As


As x~N  ,  2 
 2 
x ~N   , 
 n 

i.e

E( x )= E(  )= 

2
Var( x )=v( ̂ )=  X2 =
n
We consider

Lim P[  _  ]
n

Replacing the values

= Lim P[ x   ]
n 


Dividing by
n

 
 x     
= Lim P   
n 
   
 n n 

 
 x    n 
= Lim P   
n 
   
 n 

 n 
= Lim P  Z  
n 
  
 M.Riaz Lecturer Statistics The Islamia University of Bahawalpur RYK Campus- 6

= Lim P Z  
n 

= Lim P   Z  

n 

=1

So C-1 is satisfied

Now
 2
Now lim v( ) lim  lim 0
n n  n  n

2
=

=0

As C-1 and C-2 are satisfied so x is consistant estimator for  .

ii) it is given that

s 2

 ( x  x) 2

n 1
As we know that

E( s 2 )   2

2
s  2
2  ( x  x) 2

 n 1

s2 
2  ( x  x) 2

n 1 2
2
s 
2
 2 ( n1)
n 1
Applying variance on both sides :

4
Var( s 2 )  var (  2 ( n1) )
(n  1) 2

4 2 4
 2(n-1) 
(n  1) 2 (n  1)

Now we consider

lim P[  _  ]
n


 lim p s 2   2 
n 

 
 2 
 s  
2

 lim p   
n 
 2
4  2 4  
 n 1  
  n  1  
 M.Riaz Lecturer Statistics The Islamia University of Bahawalpur RYK Campus- 7

 
 s2   2 
 n  1 
 lim P 
n   2 4 2 2 
 
 n  1 

 n  1 
 lim p  z  
n 
 2 2 

 lim p z  
n 

lim p   z   =1
n

So condition 1 is satisfied
Now

2 4
lim v( s 2 )  lim ( )
n  n n 1

2 4
= =0


Hence C-1 and C-2 are satisfied so s 2


 ( x  x) 2

is an consistent estimator of  2 .
n 1
Q.5

Show that for a sample of size n from a normal distribution (  ,  2 ).Then show that the statistic

  xi is an consistant estimator of population parameter  .
n 1
Solution
As we know that

  xi
n 1
Applying expectation on both sides :

E ( )  E (  )
 xi
n 1
 x1  x2  x3 .....xn
E ( )  E ( )
n 1
 E ( x1 )  E ( x2 )  E ( x3 )  ...  E ( xn )
E ( ) 
n 1
      .......
E ( ) 
n 1
 n
E ( ) 
n 1

Now applying Var on 
 v x
Var(  ) 
(n  1) 2
 M.Riaz Lecturer Statistics The Islamia University of Bahawalpur RYK Campus- 8

 V ( x1  x 2  x3.....xn)
V ( ) 
(n  1) 2

 V ( x1)  V ( x 2)  V ( x3).....V ( xn)  2   2   2 ..... 2

V ( )  =
(n  1) 2 (n  1)2 _
 n 2
V ( ) 
(n  1)2

Now we consider

lim P[  _  ]
n

Replacing the values

 x n 
 lim p    
n 
 n  1 n  1 


n
Dividing both sides S.E(  )=
n 1

 x n 
  
 n 1 n 1  
 lim p   
n 
 n n 
 n 1 n 1
 
(n  1) 
 lim P[ z  ]
n 
n

 P[ z  ]

 P[  z  ]

=1
So condition 1 is satisfied
Now

n 2 2
lim v( )  lim ( )=
n  n   ( n  1) 2 

=0

Hence C-1 and C-2 are satisfied so  


 xi is a consistent estimator of population parameter 
n 1

Prepared by: Muhammad Riaz

Lecturer: Statistics
The Islamia University Bahawalpur
Sub Campus
RYK