Translate this page into:
Exponential ratio estimator of the median: An alternative to the regression estimator of the median under stratified sampling
⁎Corresponding author. subzarstat@gmail.com (Mir Subzar)
-
Received: ,
Accepted: ,
This article was originally published by Elsevier and was migrated to Scientific Scholar after the change of Publisher.
Peer review under responsibility of King Saud University.
Abstract
This article develops statistical inference about the population median under the stratified sampling method. An exponential class of ratio estimators of the median was suggested using the combination of scalars and known supplementary information on the population median. Mean square error and bias expressions were derived theoretically and also the AOE (asymptotic optimum estimator) conditions were obtained with its mean square error and bias expressions. From both the empirical evidence and analytical approach evaluations of the AOE with other obtainable members of the suggested class of estimators show that the AOE performs better than its competitors in the literature and is also the alternative to the regression estimator of the median under a stratified random sampling scheme.
Keywords
Population Median
AOE
Bias
MSE
Efficiency
Stratified Sampling
62D05

1 Introduction
Auxiliary information is commonly used in survey sampling in order to increase precision while estimating the population parameters. So, whenever this information is available and every researcher wants to utilize it in order to get more precise results. However various authors have put their sincere efforts to do the same for details see Kadilar and Cingi (2003) who amended the estimators in Upadhyaya and Singh (1999) to the sampling design stratified random sampling. Also, Singh and Vishwakarma (2008), Sharma and Singh (2015), Verma et al. (2015) suggested a new family of estimators in Stratified random sampling and Subzar et al (2018) have anticipated different estimators using Non-Conventional measures of dispersion for estimating finite population mean in stratified random sampling. Recently Lone et al (2021) have also proposed a general class of ratio estimator for estimating the finite population variance. So this type of sampling procedure is used when the population under study is heterogeneous, it is then usually used to make substrata, which are homogeneous within and heterogeneous between, with a view to producing smaller bound on the error of estimation for a fixed cost of the survey. Having been motivated by the above research works, the present study focuses on how we will get precise results from a heterogonous population and even that is skewed. However, we will get by using the linear regression estimator, but one thing is clear while using OLS (ordinary least square) our results are not precise because it is sensitive to extreme values. But our present study focuses on obtaining reliable results from the data having skewed distribution.
Consider a finite population with auxiliary variate
and study variable
which is divided into
strata containing
units in each
stratum such that
. Let
and
represent the sample medians corresponding to the population medians
and with
correlation coefficient between
and
as
. Let
and
be the marginal densities;
and
the probability density functions of the variables. Let
and
be the respective weighted sample medians corresponding to the population medians
and
where
,
where
is the sample size from stratum
and
is the total sample size. Also, let
Such that
2 Review of some existing estimators for population median in stratified sampling scheme
For the above described population different authors have proposed different estimators for population median in stratified sampling in different years whose literature is mentioned in this section, given as under.
-
The usual unbiased sample median estimator (Gross; 1980) given by with variance
The classical ratio median estimator (Kuk and Mak; 1989) given by.
The product median estimator [(Robson;1957), (Murthy;1964)] is defined as.
The exponential ratio type median estimator (Bahl and Tuteja; 1991) defined as
With bias
and Mean square error
The exponential product type median estimator (Bahl and Tuteja; 1991) defined as
With bias
and Mean square error
The chain ratio type median estimator (Kadilar and Cingi; 2003) defined as.
3 The proposed median estimator in stratified random sampling scheme
While taking the motivation from the existing estimators and suggesting the new class of Exponential Ratio estimator of the median using ancillary information and the combination of scalars. The obtainable estimators are the members of the suggested class of estimator and is given as
Where
are the scalars chosen suitably, such that
and
fulfil the condition
In order to derive the estimated expressions of bias and mean square error for the suggested class of estimator we have expressed (3.1) in terms of (1.1), then we obtain
Assuming that and expanding , and , we have
Neglecting terms of
having power greater than two, we have
Therefore,
Then, taking expectation of (3.4) and using (1.1), we obtain the bias for the proposed class of ratio median estimators
to the first degree of approximation as
While squaring equation (3.4) on both sides, ignoring terms of
having power greater than two, taking expectation, and using (1.1), we obtain the mean square error for the suggested class of ratio median estimator to the first degree of approximation as
Now to examine the optimality condition for the suggested class of ratio median estimator, let
So that
Putting (3.7) in (3.5), we have.
and .
Then
Hence substituting (3.7) in (3.6), we obtain
Thus substituting (3.2) in (3.9) we obtain the asymptotic optimum MSE for the Proposed class of ratio median estimator
as
Thus the mean square error of the suggested class of exponential ratio estimator of the median in stratified random sampling scheme at optimal condition has the same proficiency as the unbiased linear regression estimator as the mean square error expression at the optimal condition of the suggested estimator is the same as the expression of unbiased linear regression estimator.
4 Some existing members of the proposed class of estimators
How the existing estimators which are mentioned in this study fit into the suggested class of exponential ratio estimator of the median are presented in this section.
Note. As
take on definite distinctive values, the suggested class of exponential ratio estimator of the median giving the same expressions of bias and mean square error which the estimators give mentioned in section 2. Thus, the study provides unified treatment towards the properties of the existing members of the suggested class of estimators point out in this study (see for details Table 1).
Estimators
1
0
Usual Median Unbiased estimator
0
−1
−1
0
Classical ratio median estimator
0
−1
1
0
Product type median estimator
0
−1
0
−1
Bahl and Tuteja exponential ratio median estimator
0
−1
0
1
Bahl and Tuteja exponential product median estimator
0
−1
2
0
Chain ratio median estimator
5 Efficiency comparisons among members of the proposed class of estimators
Here, the mean square error of the suggested exponential ratio estimator of the median at optimum condition is equated with the mean square error of some existing median estimators.
-
Comparison of with sample median estimator .
Comparing (2.1) and (3.9), will be more efficient than
, Which is always true.
-
b)
Comparison of with classical ratio median estimator .
Comparing (2.2) and (3.9), will be more efficient than
, Which is always true.
-
c)
Comparison of with classical product median estimator .
Comparing (2.3) and (3.9), will be more efficient than
, Which is always true.
-
d)
Comparison of with classical exponential ratio median estimator .
Comparing (2.4) and (3.9), will be more efficient than
, Which is always true.
-
e)
Comparison of with classical exponential product median estimator .
Comparing (2.5) and (3.9), will be more efficient than
, Which is always true.
-
f)
Comparison of with classical chain ratio median estimator .
Comparing (2.6) and (3.9), will be more efficient than
, Which is always true.
6 Empirical study
In this present study, for verifying the general results and also check the optimality performance i.e., (asymptotic optimum estimator) for the suggested class of exponential ratio estimator of the median
over the existing members of the suggested class of estimators cited in this study, we the data of the population [Source: PDS(2012)] and the summary of the population is given in Table 2.
36
36
36
36
5
5
5
5
0.25
0.25
0.25
0.25
1480
127,289
54,559
71,615
38
3058
2033
2382
0.7776
0.8888
0.8888
0.8888
0.007056
0.00032023
0.0004219
0.0003012
0.0001641
0.000007827
0.0000141
0.00001026
3.729562
1.021175775
1.16587797
1.393809035
4.117462984
1.003722805
1.299913595
1.360970285
0.043056
0.043056
0.043056
0.043056
0.70434
0.90425
0.79715
0.91025
Considering the above population, setting
satisfies the condition in (3.2), implies
. Then set
and
so that
,
fulfils the condition in (3.7). Hereafter setting
in (3.1) we find an asymptotic optimum estimator (AOE) for the suggested class of estimators for the population median for strata 1 as
and the first degree approximation of bias and mean square error given respectively in Table 3,
For Strata 2 we have Set
satisfies the condition in (3.2), implies
. Then set
and
so that
,
fulfils the condition in (3.7). Hereafter setting
in (3.1) find an asymptotic optimum estimator (AOE) for the suggested class of estimators for the population median for strata 2 as
and the first degree approximation of bias and mean square error given respectively in Table 3,
For Strata 3, we have set
fulfils the condition in (3.2), implies
. Then set
and
so that
,
fulfils the condition in (3.7). Hereafter setting
in (3.1) we find an asymptotic optimum estimator (AOE) for the suggested class of estimators for the population median for strata 3 as
and the first degree approximation of bias and mean square error given respectively in Table 3,
and for strata 4, we have set
fulfils the condition in (3.2), implies
. Then set
and
so that
,
fulfils the condition in (3.7). Hereafter setting
in (3.1) we find an asymptotic optimum estimator (AOE) for the suggested class of estimators for the population median for strata 4 as
and the first degree approximation of bias and mean square error given respectively in Table 3,
Estimators
Bias
MSE
%RE
0.00
284303.40
100.00
16.99
64711.48
439.34
107.58
1083839.42
26.23
−7.08
102014.42
278.69
38.22
611578.41
46.49
339.72
2463319.41
11.54
−6.47
341.89
83156.59
−57.98
88186.16
322.39
−40.69
50804.88
559.60
−78.95
99681.49
285.21
−184.08
59753.60
475.79
Then, when combining all the strata, from the (6.1) to (6.12) we came to conclude that setting.
, implies
. Then set
and
so that
,
fulfils the condition in (3.7). Hereafter setting
in (3.1) we find an asymptotic optimum estimator (AOE) for the suggested class of estimators for the population median as
and the first degree approximation of bias and mean square error given respectively in Table 3,
7 Conclusion
In this study, we have suggested an exponential ratio estimator of the median in stratified random sampling, and its asymptotic conditions were also derived which shows our suggested estimator has the same expression as the unbiased linear regression estimator for mean square error. Thus our suggested exponential ratio estimator of the median using ancillary information and the combination of scalars is the general class of estimator and the existing estimators are actually the members of a proposed class of estimator and performs better at optimal conditions which we have revealed from Table 3. Thus our suggested estimator is preferred over existing estimators and is an alternative to regression estimator for practical applications in case of skewed data.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References
- Ratio and product type exponential estimator. J. Information Optimiz. Sci.. 1991;12(1):159-164.
- [Google Scholar]
- Median estimation in sample surveys. Proc. Surv. Res. Meth. Secl. Amer. Statist. Assoc. 1980:181-184.
- [Google Scholar]
- A study on the chain ratio type estimator. Hacettepe J. Math. Statist.. 2003;32:105-108.
- [Google Scholar]
- Median estimation in the presence of auxiliary information. J. Royal Statist. Soc., B. 1989;51:261-269.
- [Google Scholar]
- Enhanced estimators of population variance with the use of supplementary information in survey sampling. Math. Probl. Eng.. 2021;1:1-8.
- [Google Scholar]
- PDS (2012): Punjab Development Statistics, Bureau of Statistics, Government of the Punjab, Lahore, Pakistan
- Application of multivariate Polykays to the theory of unbiased ratio type estimation. J. Am. Stat. Assoc.. 1957;52:511-522.
- [Google Scholar]
- Generalized class of estimators for population median using auxiliary information. Hacettepe J. Math. Statist.. 2015;44(2):443-453.
- [Google Scholar]
- A family of estimators of population mean using auxiliary information in stratified sampling. Commun. Statist.: Theory Methods. 2008;37(7):1038-1050.
- [Google Scholar]
- Estimation of finite population mean in stratified random sampling using non-conventional measures of dispersion. J. Reliab. Statist. Studies. 2018;11(1):83-89.
- [Google Scholar]
- Use of transformed auxiliary variable in estimating the finite population mean. Biom. J.. 1999;41:627-636.
- [Google Scholar]
- Some families of estimators using two auxiliary variables in stratified random sampling. Rev. Investig. Operacional. 2015;36(2):140-150.
- [Google Scholar]
Appendix A
Supplementary material
Supplementary data to this article can be found online at https://doi.org/10.1016/j.jksus.2022.102536.
Appendix A
Supplementary material
The following are the Supplementary data to this article: