Journal of Reliability and Statistical Studies; ISSN (Print): 0974-8024, (Online): 2229 5666 Vol.

11, Issue 1 (2018): 45-50

RATIO AND RATIO TYPE SEPARATE ESTIMATOR FOR

POPULATION MEAN WHEN COEFFICIENT OF
VARIATION IS KNOWN

Rajesh Tailor and Asha Mishra

School of Studies in Statistics, Vikram University, Ujjain, India
E Mail: tailorraj@gmail.com; ashadubey86@gmail.com

Received April 12, 2017

Modified April 01, 2018
Accepted May 05, 2018

Abstract
This paper deals with the problem of estimation of the finite population mean using
auxiliary information. A ratio and ratio type separate estimator has been suggested for the finite
population mean. The bias and mean squared error have been derived. The suggested estimator
has been compared theoretically as well as empirically with existing estimators.

Key Words: Population Mean, Sample Mean, Stratification, Separate Type Estimator,
Separate Ratio and Ratio Type Estimator.

1. Introduction
This paper is classified into four sections. First is the introduction that discuss
the problem under consideration and relevant work done by other researchers.
Second section is the suggested estimator in which an estimator for population mean
has been suggested. Section 3 is the efficiency comparisons and in section 4
suggested estimator has been compared with other estimators empirically.

Simple random sampling has limitations of improper representation and

administrative inconveniency. In case of heterogonous population also it is not used in
general. In these type of situations, stratified random sampling is a good option where
heterogeneous population is divided into homogenous blocks called strata and a simple
random sample is drawn from each stratum. When information on parameters of
auxiliary information is known for all strata, it is considered better to use separate type
estimators for gain in efficiency. Suppose from the population S of size N, a sample of
size n is drawn using stratified sampling. Let y be the study variate and x and z are
auxiliary variates taking values y hi , x hi , z hi (h =1,2….,L; i =1,2…, N h ).

Classical combined ratio estimator in stratified random sampling was defined

and studied by Hansen et al. (1946) as
46 Journal of Reliability and Statistical Studies, June 2018, Vol. 11(1)

 L 
 ∑ Wh X h 
 = y st  X .
ˆ L
YRC = ∑ Wh y h  h L=1 x  (1.1)
h =1    st 
 ∑ W x
h h 
 h =1 
Tailor et al. (2012) studied Singh (1967) ratio-cum-product estimator in stratified
random sampling as
ˆ st  X  z st 
YRP = y st  
 Z . (1.2)
 x st  

Parmar (2012) studied a ratio-cum-product estimator for population mean in stratified

random sampling using coefficient of variation of auxiliary variate’s as

ˆ st
 L
(  L
) 
 ∑ Wh X h + C xh  ∑ Wh (z h + C zh ) 
YRPS = y st  L
h =1  h =1 . (1.3)

 ∑ h h
W (x + C )  L
xh  ∑ Wh Z h + C ( 
zh  )
 h =1  h =1 

The classical separate ratio estimator for population mean is defined as

ˆ L X 
YRS = ∑ Wh y h  h .
 (1.4)
h =1  xh 

Separate Version of Chouhan (2012) studied separate ratio type exponential

estimator for the population mean is expressed
ˆ L  X − xh 
YRe s = ∑ Wh y h exp h .
 (1.5)
h =1  Xh + x h 

Tailor and Lone (2014) proposed a separate ratio type estimator using coefficient of
variation of auxiliary variate from each stratum as

ˆC L  X + C xh

YRS = ∑ Wh y h  h .
 (1.6)
h =1  x h + C xh

ˆ ˆ ˆC
mean squared error of YRS , YRe s and YRS are
ˆ L
[
MSE ( YRS ) = ∑ Wh2 γ h S 2yh + R 12h S 2xh − 2R 1h S yxh ,
h =1
] (1.7)

ˆ L  1 
MSE ( YRe s ) = ∑ Wh2 γ h S 2yh + S 2xh R 12h − S yh S xh ρ yxh R 1h  , (1.8)
h =1  4 
and
C
MSE (YRS
L
[
) = ∑ Wh2 γ h S 2yh + θ12h R 12h S 2xh − 2R 1h θ1h S yxh ,
h =1
] (1.9)
Ratio and ratio type separate estimator for population mean ... 47

where
 1 1  Y Xh
γ h =  −  , R 1h = h and θ1h =
 nh Nh  Xh X h + C xh

2. Suggested Estimator
Using the information on coefficient of variation of auxiliary variates x and
z of each stratum i.e. C xh and C zh , suggested ratio and ratio type estimator for
population mean as

ˆC L  X + C xh  Z h + C zh 
YRRS = ∑ Wh y h  h 
 z + C
.
 (2.1)
h =1  x h + C xh  h zh 

Let us suppose that

y h = Yh (1 + e 0 h ), x h = X h (1 + e 1h ), z h = Z h (1 + e 2 h )
E(e 0 h ) = E(e1h ) = E(e 2 h ) = 0
E (e 02h ) = γ h C 2yh , E (e12h ) = γ h C 2xh , E (e 22 h ) = γ h C 2zh

E(e0h e1h ) = γ hρ yxhCxh C yh ,

E(e 0h e 2h ) = γ h ρ yzhC zh C yh ,
and
E(e1h e 2 h ) = γ h ρ xzh C xh Czh ,

In this section for a ratio and ratio type estimate has been suggested for population
mean and it’s bias and mean squared error are being derived.
Substituting the some values in (2.1), upto the first degree of approximation the bias
ˆC
and mean squared error of the suggested estimator YRRS are obtained as

L
 1
ˆC
B(YRRc )= ∑W γ h h 
X
 h
(
θ12h S xh
2
R1h − θ1h S xh S yh ρ yxh )
h =1


+
1
Zh
(
θ 22h S zh
2
R2 h + θ1hθ 2 h S xh S zh ρ xzh − θ 2 h S yh S zh ρ yzh  ,

) (2.2)

L
MSE ( yˆ RRC
C
)= ∑w γ
h =1
2
h h
 S yh

2
+ θ12h S xh
2 2
R1h + θ 22h S zh
2 2
R2 h − 2θ1h S yh S xh ρ yxh R1h

−2θ 2 h S yh S zh ρ yzh R2 h + 2θ1hθ 2 h S xh S zh ρ xzh R1h R2 h  , (2.3)

48 Journal of Reliability and Statistical Studies, June 2018, Vol. 11(1)

Xh Zh
θ1h = and θ2h =
X h + Cxh Zh + C zh

3. Efficiency Comparison
In this section suggested estimator has been compared with other considered estimator
on the basis of mean squared error.
Variance of the unbiased estimator yst is expressed as
L
V( y st ) = ∑ Wh2 γ h S2yh (3.1)
h =1
ˆC
Comparison of (1.7), (1.8), (1.9) and (3.1) exhibits that the estimator YRRc would better than

(i) Y if
L

∑W
h =1
h γh
2 θ12h S xh

2 2
R1h + θ 22h S zh
2 2
R2 h − 2θ1h S yh S xh ρ yxh R1h

−2θ 2 h S yh S zh ρ yzh R2 h + 2θ1hθ 2 h S xh S zh ρ xzh R1h R2 h  < 0, (3.2)

ˆ
(ii) YRS if
L

∑W
h =1
h γh
2  R12h S xh

2
(θ12h − 1) + θ 22h S zh
2 2
R2 h − 2 R1h S yxh (θ1h − 1)

−2θ 2 h S yh S zh ρ yzh R2 h + 2θ1hθ 2 h S xh S zh ρ xzh R1h R2 h  < 0, (3.3)

ˆ
(iii) YRe s if
L
 2 2 2 1
∑W γ
h =1
2
h h  S xh R1h (θ1h − 4 ) + θ 2 h S zh R2 h − R1h S xh S yh ρ yxh (2θ1h + 1)

2 2 2

−2θ 2 h S yh S zh ρ yzh R2 h + 2θ1hθ 2 h S xh S zh ρ xzh R1h R2 h  < 0, (3.4)

and
ˆC
(iv) YRS if
L
2
[
∑ W h γ h 2 θ 1 h θ 2 h S xh S zh ρ xzh R 1 h R 2 h − 2 θ 2 h S yh S zh ρ yzh R 2 h < 0 , ] (3.5)
h =1

4. Empirical Study
In this section suggested estimators is being compared numerically with other
considered estimators. For this purpose percent relative efficiencies of all estimators are
calculated, using a natural data set. Details of the considered data are given below
[Sources: Ministry of Agriculture] (http://agricoop.nic.in/agristatistics.html):
y:Total cereal production(Million tons),
Ratio and ratio type separate estimator for population mean ... 49

z:Total area under cultivation for cereals (million hectares)

x:Area under total cultivation of food grain(million hectares)

n 1 =4 n2 = 4 N1 =10 N 2 =10
X1 =107.51 X 2 =118.20 Z1 = 85.55 Z 2 = 95.11
Y1 = 55 Y2 =73.90 S x1 =6.29 S x 2 =2.66
S y1 =7.25 S y2 =8.02 S z1 =4.31 Sz2 =3.21
S yx1 =44.42 S xy2 =20.34 Sxz1 =3.95 S xz2 =7.07
S yz1 =30.75 S yz2 =24.09 C y =0.1318 C y2 =0.1085
1

C x1 =0.052 C x 2 =0.022 C z1 =0.050 Cz2 =0.033

ˆC
MSE (YRRS )−
MSE ( yst ) < 0 R 12h S 2xh θ12h + R 22 h S 2zh θ 22 h − 2θ1h R 1h S xyh  −4.1 < 0
 <0
− 2S yzh R 2 h θ 2 h + 2R 1h R 2 h θ1h θ 2 h S yzh 

ˆC R12hS2xh (θ12h − 1) + R 22 hS2zh θ22 h − 2(θ1h − 1)R1hSxyh 

MSE (YRRS )−
 <0 −2.0 < 0
ˆ
MSE (YRS ) < 0 − 2Syzh R 2 h θ2 h + 2R1h R 2 h θ1h θ2 hSyzh 

ˆC  2 2 2 1 
MSE (YRRS )− 2 2 2
R1hSxh (θ1h − 4 ) + R 2 h Szh θ2 h − R1hSxyh (2θ1h − 1) < 0
ˆ   −5.6 < 0
MSE (YRe s ) < 0 − 2Syzh R 2 h θ2 h + 2R1h R 2 h θ1h θ2 hSyzh 

ˆC
MSE (YRRS )−
ˆC
MSE (YRS )<0
[2R 1h R 2 h θ1h θ 2 h S yzh ]
− 2Syzh R 2 h θ2 h + R 22 hS2zh θ22 h < 0 −2.5 < 0

Table 4.1: Empirical exhibition of the theoretical conditions obtained in section 3

50 Journal of Reliability and Statistical Studies, June 2018, Vol. 11(1)

Estimators PRE’S
yst 100.00
ˆ 197.70
YRS
ˆ 137.96
YRe s
ˆC 197.64
YRS
ˆC 2107.33
YRRS

ˆ ˆ ˆC
Table 4.2: PRE’S of yst , YRS , YRe s and YRS with respect to y st

5. Conclusion
From the efficiency comparisons carried out in section 3 it is observed that the
ˆC ˆ ˆ
suggested “ratio and ratio type estimator YRRS would be more better than y st , YRS , YRe s ,
ˆC
YRS .Table 4.1 shows that conditions obtained in section 3 are satisfied empirically. Table 4.2
shows that suggested ratio and ratio type separate estimator has the maximum percent relative
ˆ ˆ ˆC
efficiency as compared to y st , YRS , YRe s and YRS and. Thus it can be concluded that the
ˆC
suggested estimator YRRS is most efficient and recommended for estimating the population
mean.

Acknowledgment
Authors have deep sense of thanks to referees for their support in improving the overall
quality of the paper.

References
1. Hansen, M.H., Hurwitz, W.N. and Gurney, M. (1946). Problem and methods
of the sample survey of business. J. Amer. Statist. Assoc., 41, p. 173-189.
2. Singh, M. P. (1967). Ratio-cum-product method of estimation. Metrika, 12, p.
34-42.
3. Tailor, R., Tailor, R., Parmar, R., and Kumar, M.(2012). Dual to ratio-cum-
product estimator using known parameters of auxiliary variable. Journal of
Reliability and Statistical Studies, 5(1), 65-71.
4. Lone,H.A. and Tailor,R. (2014). Dual to separate ratio type exponential
estimator in post-stratification, J. Statist. Appl. Prob. 3(3), p. 425−432.
5. Chouhan, S. (2012). Improved estimation of parameters using auxiliary
information in sample surveys, Ph.D Thesis, Vikram University, Ujjain,
M.P.,India.
6. Ministry of Agriculture [ Website: (http://agricoop.nic.in/agristatistics.html)].