You are on page 1of 5

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 13 (2018) pp.

11151-11155
© Research India Publications. http://www.ripublication.com

A Class of Double Sampling Log Type Estimators for Population Variance


Using Two Auxiliary Variable

Shashi Bhushana and Chandni Kumarib


a
Department of Mathematics & Statistics, Dr. Shakuntala MishraNational Rehabilitaion University, Lucknow, 226017, India.
b
Department of Applied Statistics, Bahasaheb BhimraoAmbedkar University, Lucknow, 226025, India.

Abstract two auxiliary variables respectively of size n ' .


In this paper, a class of log-type estimator using two auxiliary
information is proposed. Double sampling technique has been Here, y , x1 and x2 be the sample means of study variable
considered as it is assumed that the auxiliary information and two auxiliary variables from the sample of size n . Also,
about the auxiliary variable is unknown. Bias and mean N 2 n 2
squared error has been found up to the first order of
approximation. The proposed classes are compared to some
s N
2
y
1
 y  y 
i 1
i , s n
2
x1
1
 x  x 
i 1
i 1 and

commonly used estimators both theoretically as well as n 2


empirically and they perform better than commonly known
estimators available in the literature.
s n
2
x2
1
 x  x 
i 1
i 2 be the sample variance of the study

and two auxiliary variables respectively.


INTRODUCTION
The use of auxiliary information enhance the precision of an THE SUGGESTED GENERALIZED CLASS OF LOG-
estimator. Here, two auxiliary information is used for TYPE DOUBLE SAMPLING ESTIMATORS
estimating the population variance under double sampling.
We propose the following new classes of log type estimators
We select the sample in two phases for estimating the 2
population variance. Various authors used multiple auxiliary for the population variance S y as
information (Olkin (1958), Raj (1965), Singh (1967),
a1 a2
Shukla(1966), etc.). Recently, Bhushan and Kumari (2018)   sx2'1     sx2'2 
had made the use of logarithmic relationship between the T1  s y 1  log  2  
2
1  log  2  
  sx     sx
 
study variable and auxiliary variable for estimating the
population variance.  1   2
(2.1)
Consider a finite population U  U1 ,U 2 ,..., U N of size
N from which a large sample of size n ' is drawn according   sx2'     sx2' 
T2  s y2 1  b1 log  21   1  b2 log  22  
to simple random sampling without replacement (SRSWOR)
  sx     sx
for estimating unknown auxiliary variables only. Then, we
 1    2  
select a sample of size n from the remaining observation for (2.2)
estimating the sample mean of study and auxiliary variables. c1 c2
  sx2*'1     sx2*'2  
Let yi , xi 1 and xi2 denotes the value of the study and two T3  s 1  log  2*  
2
1  log  2*  

y
 sx     sx  
auxiliary variable for the ith unit i  1, 2,..., N of the  1   2 
(2.3)
population. Let Y , X 1 and X 2 be the population means of
study variable and two auxiliary variables. Also,   sx2*'     sx2*'  
2 2 T4  s y2 1  d1 log  2*1   1  d 2 log  2*2  
N n
 sx     sx  
S y2  N 1   yi  Y  , S x21  n1   xi  X 1  and   1    2 
i 1 i 1 (2.4)
n 2

S x22  n1   xi  X 2  be the population variance of the where sx2*'i  ai sx2'i  bi and sx2*i  ai sx2i  bi for i = 1, 2
i 1
study and two auxiliary variables respectively. such that ai , bi , ci and d i are optimizing scalars or functions
Further, x1 ' and x2 ' be the larger sample means of two of the known parameters of the auxiliary variable xi ' s such
as the standard deviations S x , coefficient of variation C x ,
auxiliary variables and sx2'1 , sx2'2 be the sample variance of i i

11151
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 13 (2018) pp. 11151-11155
© Research India Publications. http://www.ripublication.com

coefficient of kurtosis b2 x , coefficient of skewness b1x and


i i E  2 2'   I ' b2*x2 , E  01   I I 22
*
y x1 ,
correlation coefficient rx x of the population  i  j  0  . E  01 '  I ' I 22 y x1 , E   0 2   I I 22 y x2 ,
* *
i j

E  0 2 '  I ' I 22
*
y x2 , E  1 2   I ' I 22 x1 x2 and
*

PROPERTIES OF THE SUGGESTED CLASS OF E 1 '  2 '  I ' I 22


*
x1 x2 where b2 y  b2 y  1 , b2 x1  b2 x1  1
* *

ESTIMATORS
, b2 x  b2 x  1 and I 22 y x  I 22 y x  1 ,
* *
In order to obtain the bias and mean square error (MSE), let us 2 2 1 1

y x2  I 22 y x2  1 , I 22 x1 x2  I 22 x1 x2  1 ;
* *
consider I 22

0 
s y
2
 Sy2 
, 1 
s x1
2
 S x1 2  ,   s ' x1
2'
 S x1 2 , I pq  mpq m20p /2 m02
q /2
,
N
m pq   Yi  Y  X X
1
Sy2 S x1 2 S x1 2 p q
N , I 1 N ,
i

2 
s x2
2
 S x2 2
,  '
s x2
2'
 S x2 2
 i 1

I '  1 n ' , b2 y  m40 m20


2
, b2 x  m04 m02 are the
2
2 2 2
S x2 S x2
coefficient of kurtosis of y and x respectively.

E  0   E 1   E  2   0 , E  0 2   Ib2* y ,
E 12   Ib2x
*
1
, E  1   I ' b2 x , E   2   Ib2x ,
2' *
1
2 *
2

Theorem 1
The bias and the mean squared error of the proposed estimator considered up to the terms of order n−1 are given by

 a12 * a22 * 
Bias T1   S  I  I '  b2 x1  b2 x2  a1 ry x1 b2* y b2*x1  a2 ry x2 b2* y b2*x2  a1 a2 rx1x2 b2*x1 b2*x2 
2
y
2 2 
4 

* 2 * 2 * * *

MSE T1   S y Ib2 y   I  I ' a1 b2 x1  a2 b2 x2  2a1ryx1 b2 yb2 x1  2a2 ryx2 b2 yb2 x2  2a1a2rx1x2 b2*x1b2*x2 
* *
 
  a2 a2  
2w1 S y4 1   I  I '  1 b2*x1  2 b2*x2  a1 ry x1 b2* y b2*x1  a2 ry x2 b2* y b2*x2  a1 a2 r x1x2 b2*x1 b2*x2 
  2 2  
* * *
I 22 I 22 I 22
where ry x  , ry x  and rx x 
y x1 y x2 x1 x2
1 2 1 2
b2* y b2*x1 b2* y b2*x2 b2*x1 b2*x2
Proof. Consider the estimator
a1 a2
  sx2'1     sx2'2 
   S y2 1   0  1  log 1  1'  1  1   1  log 1   2'  1   2  
1 a1 1 a2
T1  s 1  log  2   1  log  2
2
y
 sx     sx
  1    2  

 a12 ' 2
 S 1   0  1  a1 1  1  11  1   1  1   a1 1'  1  
2 ' ' 2 2
y
 2 
 a22 ' 2
1  a2  2   2   2 2   2    2   2   a2  2   2  
' ' 2 2 '
(3.1)
 2 
On solving and then taking expectation on both the sides, we get

 a2 a2 
Bias T1   S y2  I  I '  1 b2*x1  2 b2*x2  a1 ry x1 b2* y b2*x1  a2 ry x2 b2* y b2*x2  a1 a2 rx1x2 b2*x1 b2*x2 
2 2 

11152
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 13 (2018) pp. 11151-11155
© Research India Publications. http://www.ripublication.com

Squaring and then taking MULTIVARIATE EXTENSION OF PROPOSED CLASS


OF ESTIMATORS
expectation on both the sides of equation (3.1), we get
required expression for MSE. Let there are k auxiliary variables then we can use the
variables by taking a linear combination of these k
estimators of the form given in section 2, calculated for every
Corollary 1.The optimum values of constant are obtained as
auxiliary variable separately, for estimating the population
variance. Then the estimators for population variance will be
 ryx  ryx2 rx1x2  b2* y defined as
a1opt  1  *
 1  rx1x2
2
 b2 x1 ai
  sx2'i  
T1  s y  i 1 1  log  2  
* 2 k

 ryx2  ryx1 rx1x2  b2* y   sx  


a2 opt   *  i 
 1  rx1x2
2
 b2 x2
  sx2'i  
T  s  i 1 1  bi log  2  
* 2 k
The optimum mean squared error is given by 2 y
  sx  
 i 
M T1 opt  S y4  I   I  I ' Ry2*. x1x2 
ci
(3.2)   sx2*'  
T3*  s y2  i 1 1  log  2*i  
k

where Ry2*.x1x2 is the transformed multiple correlation   sx  


 i 
coefficient between y and x1 , x2
  sx2*'i  
T  s  i 1 1  di log  2*  
* 2 k
4 y
  sx  
 i 
where ai , bi , ci and d i are the optimizing scalars i = 1, 2,
..., k.

Theorem 2
The bias and the mean squared error of the proposed estimator considered upto the terms of order n −1 are given by

 k ai2 * 
Bias T   S  I  I '  b2 xi   ai ry xi b2* y b2*xi   ai a j rxi x j b2*xi b2*xj 
k k
* 2
1 y
 i 1 2 i 1 i  j 1 
  
MSE T1*   S y4  Ib2* y   I  I '  ai2b2*xi  2 ai ryxi b2* yb2*xi  2  ai a j rxi x j b2*xi b2*x j 
k k k

  i 1 i 1 i  j 1 
  k a2 k k  
2w1 S y4 1   I  I '   i b2*xi   ai ry xi b2* y b2*xi   ai a j rxi x j b2*xi b2*x j  
  i 1 2 i 1 i  j 1  

EFFICIENCY COMPARISON General variance estimator


In this section, we compare the proposed classes of estimators
Sˆ y2  s y2
with some important estimators. The comparison will be in
terms of their MSE up to the order of n−1. The optimum mean It’s mean squared error is given by
squared error of proposed estimator is given by
MSE ( Sˆ y2 )  S y4 I b2* y
M T1 opt  S y4  I   I  I ' Ry2*. x1x2 
The usual ratio type variance estimator

 sx2'  sx2' 
Sˆr2'  s y2  21  22 
 sx  sx
 1  2 

11153
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 13 (2018) pp. 11151-11155
© Research India Publications. http://www.ripublication.com

It’s mean squared error is given by EMPIRICAL STUDY

 Ib2* y   I  I ' b2*x1  b2*x 2  The data on which we performed the numerical calculation is
MSE ( Sˆr2' )  S y4    MSE (T1 )opt taken from some natural populations. The source of the data is
  2 I 22 yx1  2 I 22 yx2  2 I 22 x1x2 
* * *
given as follows.

Population 1. (Chochran, Pg. no. 155). The data concerns


The product type variance estimator
about weekly expenditure on food per family.
 sx21  sx22  y : weekly expenditure on food
ˆ
S p  s y  2'  2'
2' 2

 sx  sx
 1  2  x1 : number of persons
Its mean squared error is given by
x2 : the weekly family income
 Ib   I  I ' b  b  2I
* * * *

MSE ( Sˆ p2' )  S y4  Population 2. (Choudhary F. S., Pg. no. 117).
2y 2 x1 2 x2 22 yx1
  MSE (T1 )opt
  2 I 22* yx2  2 I 22* x1x2  y : area under wheat (in acres) in 1974
Singh, Chauhan, Sawan and Smarandache (2011) type x1 : area under wheat (in acres) in 1971
variance estimator
x2 : area under wheat (in acres) in 1973
 sx2'1  sx21  sx2'2  sx22 
ˆ
S  s exp  2' 2 
 sx2'  sx2 
2' 2
s
 sx  sx
y
 1 1  2 2 
The summary and the percent relative efficiency of the
It’s mean squared error is given by following estimators are as follows:
  b2*x1 b2*x 2 
    I 22*
yx1  

 Ib*   I  I '  4 4  Table 2: Parameters of the data
MSE ( Sˆs2' )  S y4   MSE (T1 )opt
 2y
 * I 22 x1x2 
*

  I 22 yx2  4 
  Parameter Population 1 Population 2

Olufadi and Kadilar (2014) variance estimator


a1 a2 N 33 34
 sx2'   sx2'2 
Sˆ  s  21 
2' 2
 2 
K
 sx 
y
n' 29 30
 1  sx2 
It’s mean squared error is given by
n 11 10

MSE (SˆK2' )  MSE (T1 )opt b2* y 4.032 2.725

b2*x1 1.388 12.366


Das and Tripathi (1978) type variance estimator
b2*x1 1.143 1.912
 sx2'1  sx2'2 
ˆ 
S D  s y 2'
2' 2
 
 sx  a1 sx2  sx2'
 1 1 1
   2 
 sx2'  a2 sx2  sx2'
2 2
 

*
I 22 yx1
0.305 0.224

It’s mean squared error is given by *


I 22 1.155 2.104
yx2

MSE (SˆD2' )  MSE (T1 )opt * 0.492 0.152


I 22 x1 x2

11154
International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 13 (2018) pp. 11151-11155
© Research India Publications. http://www.ripublication.com

REFERENCES
Table 3: PRE of the estimators
[1] Bhushan S. and Kumari C. (2018). A new log type
estimators for estimating the population variance, Int.
Estimator Pop. 1 Pop. 2 J. Comp. App. Math., 13 (1), 43-54.

Sˆ y2
[2] Bahl S. and Tuteja R. K. (1991). Ratio and Product
100 100
type exponential estimator, Info. Optim. Sci., Vol.
XII(I), 159-163.
Sˆr2' 91.627 29.173 [3] Bhushan S., Gupta R. and Pandey S. K. (2015).
Some log-type classes of estimators using auxiliary
Sˆ p2' 50.236 17.524 information.
[4] Hidiroglou M. A. and Sarndal C. E. (1998). Use of
Sˆs2' 109.836 75.636 auxiliary information for two-phase sampling,
Survey Methodology, 24(1), 11-20.

SˆD2' 122.595 230.718 [5] Neyman J. (1938). Contribution to the theory of


sampling human populations, J. Amer. Stat. Asso.,
33, 101-116.
SˆK2' 122.595 230.718
[6] Cochran W. G. (1963). Sampling Techniques, Wiley
Eastern Private Limited, New Delhi, 307-310.
T1opt 122.595 230.718
[7] Chaudhury A. (1978). On estimating the variance of
a finite population. Metrika, 25, 66-67.
[8] Das A. K. and Tripathi T. P. (1978). Use of auxiliary
CONCLUSION information in estimating the nite population
variance. Sankhya, C(4), 139-148.
This work utilizes the two auxiliary information for estimating
the study variable under double sampling. It is clear from the [9] Gupta S. and Shabbir J. (2008). Variance estimation
comparative study and numerical study that the proposed in simple random sampling using auxiliary
estimator perform better than conventional estimators viz. information, Hacettepe Journal of Mathematics and
variance estimator, ratio type estimator, product type Statistics, 37, 57-67.
estimator and found equally efficient to Das & Tripathi type [10] Isaki C. T. (1983). Variance estimation using
(1978) estimator, Olufadi and Kadilar type (2014) estimator Auxiliary Information, Jour. Amer. Statist. Asssoc.,
etc. Hence, the proposed estimators have much more practical 78, 117-123.
utility than the conventional estimators.
[11] Kadilar C. and Cingi H. (2006)a . Improvement in
variance estimation using auxiliary information,
Hacettepe Journal of Mathematics and Statistics,
1(35), 111-115.
[12] Kadilar C. and Cingi H. (2006)b . Ratio estimators
for population variance in simple and stratied
sampling, Applied Mathematics and Computation,
1(73), 1047-1058.
[13] Sukhatme P. V., Sukhatme B. V., Sukhatme S. and
Ashok C. (1984). Sampling Theory of Surveys with
Applications, Iowa State University Press, Ams.
[14] Swain A. K. P. C. and Mishra G. (1994). Estimation
of population variance under unequal probability
sampling, Sankhya, B (56), 374-384

11155

You might also like